Indonesian J our nal of Electrical Engineering and Computer Science V ol. 42, No. 1, April 2026, pp. 62 ∼ 70 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v42.i1.pp62-70 ❒ 62 Enhanced image compr ession thr ough h ybrid stagger ed do wnsampling and DCT Benlab bes Haouari 1 , Khair Y ounes 2 , Beladgham Mohammed 3 , El Hendi Hichem 2 1 Department of Exact Sciences, Higher Normal School of Bechar , Bechar , Algeria 2 Department of Exact Sciences, T ahri Mohammed Uni v ersity Bechar , Bechar , Algeria 3 Information Information Processing and T elecommunication Laboratory (L TIT), Department of Electrical Engineering, T ahri Mohammed Uni v ersity Bechar , Bechar , Algeria Article Inf o Article history: Recei v ed Jun 21, 2025 Re vised Jan 30, 2026 Accepted Mar 4, 2026 K eyw ords: Discrete cosine transform Hybrid compression Image compression SPIHT+D WT Staggered do wnsampling ABSTRA CT Image compression is crucial for multimedia applications with the aim of reduc- ing storage and/or transmission costs, while preserving reliable visual quality . In this research, we propose a no v el h ybrid image compression technique based on staggering do wnsampling combined with discrete cosine transform (DCT). The propos ed approach not only o v erlaps do wnsample images to reduce data re- dundanc y b ut also utilizes the ener gy compaction properties of DCT for ef cient compression. The proposed method performance on benchmark grayscale im- ages such as Lena, House, and other refernce images were e v aluated by means of image quality assessment metrics, namely , peak signal-t o-noise ratio (PSNR), structural similarity inde x (SSIM), visual information delity (VIF); and com- pression ef cienc y met rics: bitrate and compre ssion ratio. The results clearly sho w that the proposed algorithm outperforms JPEG and Set partitioning in hi- erarchical trees (SPIHT) + discrete w a v elet transform (D WT) method, with the follo wing results: P SNR of 45.02, MSSIM of 0.9856, VIF of 0.8271, Bitrate of 0.12 bpp and a Compression Ratio of 64.00 (i.e. a reduction of 64 times) . The suggested h ybrid image compression method optimizes b ug multimedia stor - age and transmission by minimizing stor age space and bandwidth usage while maintaining image quality . It, therefore, achie v es a balance between percep- tual quality and compression ef cienc y , making it the best option for resource- constrained applications such as remote sensing, embedded systems, video com- pression, and medical image archi v al. This is an open access article under the CC BY -SA license . Corresponding A uthor: Benlabbes Haouari Department of Exact Sciences, Higher Normal School of Bechar Bechar , 08000, Algeria Email: hbenlabbes@yahoo.fr 1. INTR ODUCTION The rapid adv ancements in digital imaging technologies has res ulted in an e v er -increasing demand for ef cient image compression methods in v arious elds, mainly medical imaging, remote sensing, multimedia applications, and real-time communications. High-resolution images and videos require a great deal of storage space and a lot of bandwidth. In an ntshell, image compression has become an important aspect of modern digital systems. The ideal image compression algorithm results in a high compression ratio while also pro viding ef cient perceptual quality that guarantees visual quality by yielding important visual detail. T raditional image compression algorithms that use image compression methods lik e JPEG [1], [2] and SPIHT+D WT [3], [4], are J ournal homepage: http://ijeecs.iaescor e .com Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 63 emplo yed due to their simplicity , lo w computational-comple xity and widespread usage. JPEG based on the discrete cosine transform (DCT) is the most commonly used lossy compression standard, ho we v er , it produces blocking artif acts and can cause the loss of ne detail, especially when compression ratios are ele v ated. Set partitioning in hierarchical trees (SPIHT) [5] has pro v ed its ef ci enc y in impro ving compression ef cienc y and scalability , when combined wi th the discrete w a v elet transform (D WT). Nonetheless, SPIHT+D WT has its pitf alls too; it o v ersamples ne details and creates ringing artif acts near edges, re sulting in a lo w percei v ed visual qualit y , and it is widely kno wn that the tradeof f between compression ef ci enc y and visual delity i s the e v er -present dilemma in image compression research. T o o v ercome these challenges, h ybrid methods that implement spatial preprocessing combined with transform coding ha v e been e xamined. Staggered do wnsampling, or quincunx subsampling, is a potentially ef- cient method for eleminati ng data redundanc y while preserving salient st ructural information [6]. Compared uniform sampling, staggered do wnsampling ef fecti v ely retains the image geometric structure by only sampling pix els in a staggered manner . The method operates using spatial correlat ion between neighboring pix els to reduce reduced data size suf ciently while pro viding minimal de gradation. T e xture coding with DCT results in a solid h ybrid compression when utilized with staggered do wnsampling. The DCT is usually utilized in image and video compression due to its ener gy compaction because most of the ener gy of the signal is concentrated on a fe w lo w-frequenc y coef cients [7]-[9]. This property is ef cient since the DCT may be quantized and subsequently entrop y-coded [10] to result in maximum compression performance. Coupling staggered do wn- sampling and DCT not only decreases data rates, b ut also maint ains enough delity to minimize artif acts while achie ving good compression performance through high compression ratios. This study thoroughly assesses the ef cienc y of the proposed h ybrid image compression method on benchmark data sets. Its results are compared with widely used compression techniques, including JPEG, SPIHT+D WT , and w a v elet-based compression, to pro v e the ef cienc y of staggered do wnsampling combined with DCT . The ndings illustrate the clear balance between compression ef cienc y and visual quality pro vided by this approach. Additionally , this s tudy of fers v aluable potential to achie v e the trade-of f between com- pression ratio and perceptual accurac y . It therefore, contrib utes to the de v elopment of more adv anced image compression algorithms for v arious real-w orld applications. 2. PR OPOSED METHOD The h ybrid image compression approach proposed in this study combines staggered do wnsampling with the DCT to reach a balance between compression ratio, perceptual quality , and computati onal comple xity at an ef fecti v e le v el. W ithin this frame w ork, spatial preprocessing minimizes local redundanc y while transform- based coding utilizes frequenc y-domain compaction, therefore enabling more ef cient representation of image content. T o trace the progression of the image processing pipeline in Figure 1, the outline belo w represents step from input to output during the compression phase. Belo w are the general steps that illustrate the proposed method: − Preprocessing: con v erting image to grayscale for simplicity . − Staggered do wnsampling: reducing image resolution to achie v e compression at the e xpense of losing some information. − Block Di vision: di viding the image into smaller blocks (8x8) for DCT processing. − DCT : applying DCT to each block to reach the frequenc y domain and compress the image, in other w ords, quantizing the DCT coef cients and reducing precision to achie v e greater compression. − Reconstruction: applying IDCT to the quantized coef cients to reconstruct each block from the frequenc y domain back to the spatial domain. − Interpolation: resizing the image to its original size by means of bilinear interpolation to counteract the do wnsampling. − Ev aluating the compression quality (PSNR [11]-[16], MSSIM [17]-[18], FIV [19], [20]) and Storage Met- rics (Bitrate [21], [22] and compression ratio [23]-[25]) and displaying both the original and resulting images for comparison. Enhanced ima g e compr ession thr ough hybrid sta g g er ed downsampling and DCT ... (Benlabbes Haouari) Evaluation Warning : The document was created with Spire.PDF for Python.
64 ❒ ISSN: 2502-4752 Input Image Loading & Preprocessing Staggered (Quincunx) Do wnsampling 8 × 8 Block Di vision 2-D DCT & Quantization Dequantization & IDCT Interpolation Performance Ev aluation (PSNR, MSSIM, VIF , Bitrate, CR) Figure 1. Flo wchart of the proposed h ybrid image compression method 2.1. Stagger ed do wnsampling Also kno wn as check erboard or quincunx sampling [7], [8], staggered do wnsam pling structures image pix els along a diagonal grid, thereby preserving important structural information such as edges and te xtures, while minimizing resolution. This method is particularly helpful for image compression and multi-resolution analysis, as it preserv es perceptual quality while reducing data redundanc y . The process comprises three main stages: − Pix el Selection: Pix els selected in a staggered manner establish a quincunx pat tern in which e v ery other pix el in a 2×2 neighborhood is k ept. In an 8×8 block, for e xample, only a reduced number of pix els is k ept, yet important detail is maintained. − Reduction in Resolution: Eesolution is diminished in both horizontal and v ertical directions, usually leading to an image reduced by a f actor of four (2×do wnsampling in both directions). − Structural Information and Related Quality: The method ef ciently preserv es k e y structural information, hence, retaining important visual information in the image, therefore increasing the quality of the com- pressed and reconstructed image. F or e xample, in the case of an image I ( x, y ) of size M × N , the staggered do wnsampled image I d ( x ′ , y ′ ) is obtained by sorting one pix el from each 8 × 8 block as sho wn in Figure 2. The transformation can be mathematically e xpressed as follo ws: I d ( x ′ , y ′ ) = I (8 x, 8 y ) (1) where: • x ′ = 0 , 1 , . . . , M 8 − 1 • y ′ = 0 , 1 , . . . , N 8 − 1 This process deceases image resolution by a f actor of 8, meaning that the resulting image com p r izes only 1 8 th of the initial data. Indonesian J Elec Eng & Comp Sci, V ol. 42, No. 1, April 2026: 62–70 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 65 Figure 2. Check erboard sampling: One pix el selected per 8 × 8 block (black dot) 2.1.1. Discr ete cosine transf orm The DCT is a mathematical technique that con v erts a signal or an image from the spatial domai n into the frequenc y domain [8], [9], e xpressing a ni te set of data samples as a combination of cosine w a v es with v arying frequencies. In the eld of image processing, the DCT is commonly emplo yed for compression purposes, as it represents the signal or image ef ciently since most of the image’ s ener gy is concentrated in a limited number of lo w-frequenc y components, which are perceptually more important. 3. RESUL TS AND DISCUSSION The proposed approach w as e v aluated using well-kno wn benchmark grayscale images such as Lena, House, and others, and w as compared with established compression techniques including JPEG and SPIHT combined with the D WT . T o e v aluate its performance, se v eral objecti v e assessment metrics were emplo yed, namely peak signal-to-noise ratio (PSNR), Structural similarity inde x measure (SSIM), visual information delity (VIF), bit rate (bits per pix el – bpp), and compression ratio. These metrics of fer a more comprehensi v e vie w of reconstruction quality and coding ef cienc y by capturing pix el-le v el accurac y , perceptual resemblance, and compression performance across the e v aluation criteria. The bit rate indicates the quantity of information preserv ed per pix el after compression, whereas the compression ratio reects the e xtent to which the original data size has been reduced. The T ables 1, 2, and 3 present a summary of the results obtained by the proposed method in compari- son with JPEG and SPIHT+D WT , using the image quality metrics PSNR (T able 1), MSSIM (T able 2), and VIF (T able 3) across a set of benchmark images. T able 1 indicates that the proposed algorithm consistently achie v es higher PSNR v alues than JPEG and SPIHT+D WT across all test images. PSNR is e xpressed in decibels (dB) and functions as an objecti v e indicator of image reconstruction qual ity , where higher v alues signify reduced distortion and closer similar - ity to the original image. F or instance, the “Lena” image attains a PSNR of 38.78 dB using the proposed approach, compared with 35.96 dB for SPIHT+D WT and 35.81 dB for JPEG. Substantial impro v ements are also recorded for the “House” image (45.02 dB) and the “Cameraman” image (42.44 dB), and these results highlight the ef fecti v eness of the proposed algorithm in reconstructing images while preserving structural and te xtural information. Ev en in the case of comple x and highly te xtured i mages such as “Mandril” (34.15 dB) and “Pirate” (37.58 dB), which pose challenges to con v entional compression methods due to their high-frequenc y richness, the proposed algorithm surpasses both SPIHT+D WT (33.17 dB) and JPEG (27.89 dB) in terms of o v erall reconstruction quality . The results presented in T able 1 re v eal the capability of the proposed h ybrid scheme to ef ciently capture and reconstruct ne image details by le v eraging principles from both the spatial and frequenc y domains. In T able 2, the MSSIM comparison among the proposed method, JPEG, and SPIHT+D WT is pre- sented. The proposed approach consistently achie v ed the highes t MSSIM v alues for all tested images, illus- trating its ef fecti v eness in combining perceptual and structural image quality . F or e xample, the “House” image attained an MSSIM of 0.9856 with the proposed method, whereas JPEG achie v ed 0.9749 and SPIHT+D WT Enhanced ima g e compr ession thr ough hybrid sta g g er ed downsampling and DCT ... (Benlabbes Haouari) Evaluation Warning : The document was created with Spire.PDF for Python.
66 ❒ ISSN: 2502-4752 reached 0.9473. Ev en for structurally comple x images with rich te xtures, such as “Mandril” and “Peppers, ” the proposed scheme produced MSSIM v alues of 0.9555 and 0.9297, compared with JPEG v alues of only 0.8693 and 0.8109. These results pro vide strong e vidence, in line with the PSNR ndings, that the proposed scheme deli v ers high numerical accurac y and visually superior quality , making it particularly suitable for applications prioritizing structural preserv ation and human perceptual delity . T able 1. PSNR (dB) comparison for the proposed method, JPEG and SPIHT+D WT Image Proposed method JPEG SPIHT+D WT Lena 38.78 35.81 35.96 House 45.02 42.13 36.87 Lak e 36.58 33.16 34.42 Jetplane 40.03 36.58 36.62 Mandril 34.15 27.89 33.17 Li vingroom 37.19 33.34 34.93 Peppers 34.67 28.09 33.97 Pirate 37.58 33.29 35.39 Cameraman 42.44 38.88 37.76 W alkbridge 35.69 30.17 33.86 T able 2. MSSIM comparison for the proposed method, JPEG and SPIHT+D WT Image Proposed method JPEG SPIHT+D WT Lena 0.9467 0.9191 0.9060 House 0.9856 0.9749 0.9473 Lak e 0.9409 0.8970 0.9066 Jetplane 0.9650 0.9446 0.9297 Mandril 0.9555 0.8693 0.9385 Li vingroom 0.9510 0.9086 0.9174 Peppers 0.9297 0.8109 0.9063 Pirate 0.9540 0.9080 0.9252 Cameraman 0.9781 0.9633 0.9392 W alkbridge 0.9658 0.9011 0.9454 T able 3 pro vides a comparati v e assessment of the VIF metric for the proposed method, JPEG, and SPIHT+D WT . VIF e v aluates the amount of visual information retained in the compressed image relati v e to the reference im age, making it a rob ust perceptual quality measure. The propose d technique consist ently achie v es the highest VIF v alues acros s all test images, indicating superior preserv ation of perceptual content. F or e x- ample, t he “House” image attains a VIF of 0.8271 with the proposed approach, which e xceeds the v alues for JPEG (0.7521) and SPIHT+D WT (0.6976). Ev en for highly te xtured and structura lly comple x images such as “Mandril” and “Peppers, ” the proposed method deli v ers VIF v alues of 0.6176 and 0.6288, signicantly outper - forming both JPEG and SPIHT+D WT . These results demonstrate that for applications requiring high-quality image reconstruction, the proposed approach preserv es visual delity and informati v e content ef fecti v ely after compression. T ables 4 and 5 compare the bit rate (bits per pix el) and compression ratio for the proposed approach, JPEG, and SPIHT+D WT . The proposed scheme maintains a remarkably lo w and consistent bit rate of 0.12 bpp across all e v aluated images, corresponding to a uniform and high compression ratio of 64:1. This consistenc y highlights the rob ustness and v ersatility of the method for a di v erse range of images, from relati v ely smooth ones lik e “Lena” to highly te xtured images lik e “Mandril. ” In contrast, JPEG and SPIHT+D WT e xhibit sub- stantially higher bit rates and correspondingly lo wer compression ratios. F or instance, for the “Mandril” image, JPEG achie v es a compression ratio of 5.67:1, and SPIHT+D WT reaches 1.86:1, whereas the proposed method maintains a 64:1 compression ratio. These ndings illustrate the ef cienc y of the proposed coding technique in signicantly reducing both storage and transmission requirements while pres erving perceptual quality met- rics such as MSSIM and VIF , making it particularly suitable for bandwidth-limited and resource-constrained scenarios. Indonesian J Elec Eng & Comp Sci, V ol. 42, No. 1, April 2026: 62–70 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 67 T able 3. VIF comparison for the proposed method, JPEG and SPIHT+D WT Image Proposed method JPEG SPIHT+D WT Lena 0.6788 0.5923 0.5850 House 0.8271 0.7521 0.6976 Lak e 0.6605 0.5583 0.5879 Jetplane 0.7154 0.6146 0.6141 Mandril 0.6176 0.4362 0.5791 Li vingroom 0.6732 0.5586 0.5936 Peppers 0.6288 0.4492 0.5976 Pirate 0.6804 0.5573 0.6028 Cameraman 0.7665 0.6590 0.6632 W alkbridge 0.6778 0.5257 0.6117 T able 4. Bitrate (bpp) comparison for the proposed method, JPEG and SPIHT+D WT Image P roposed method JPEG SPIHT+D WT Lena 0.12 0.64 2.30 House 0.12 0.45 2.06 Lak e 0.12 0.89 2.77 Jetplane 0.12 0.66 2.32 Mandril 0.12 1.41 4.31 Li vingroom 0.12 0.89 2.74 Peppers 0.12 1.11 3.52 Pirate 0.12 0.86 2.62 Cameraman 0.12 0.57 2.17 W alkbridge 0.12 1.24 3.52 T able 5. Compression ratio comparison for the proposed method, JPEG and SPIHT+D WT Image Proposed JPEG SPIHT+D WT Lena 64:1 12.53:1 3.48:1 House 64:1 17.70:1 3.88:1 Lak e 64:1 9.03:1 2.89:1 Jetplane 64: 1 12.05:1 3.45:1 Mandril 64:1 5.67:1 1.86:1 Li vingroom 64:1 8.97:1 2.92:1 Peppers 64:1 7.19:1 2.27:1 Pirate 64:1 9.29:1 3.05:1 Cameraman 64:1 13.99:1 3.68:1 W alkbridge 64:1 6.47:1 2.27:1 As sho wn in the preceding tables, the e xperimental results for the images used in this study indi cate that the proposed compression method outperforms both JPEG and SPIHT+D WT in all tests. The proposed ap- proach achie v es the highest v alues for objecti v e quality metrics such as PSNR, MSSIM, and VIF , demonstrating its superior ability to preserv e image details and structural information. In terms of bit rate and compression ratio, the proposed method also signicantly surpasses the other techniques, attaining a uniform bit rate of 0.12 bpp and a compression ratio of 64:1 across all images, while JPEG and SPIHT+D WT reach bit rates of 0.53 bpp and 0.62 bpp, respecti v ely . These results conrm that the proposed method ef fecti v ely reduces storage and transm ission requirements while minimizing de gradation in visual quality , of fering an ef cient solution for abstract image compression. Figures 3 to 6 illustrat e the perceptual quality of reconstructed images from se v eral test cases using JPEG, SPIHT+D WT , and the proposed method, highlighting dif ferences observ able to the human e ye. The results of this study demonstrate that inte grating staggered do wnsampling with block-based DCT pro vides an ef fecti v e balance between compression ratio, perc eptual quality , and computational ef cienc y . In comparison with traditional DCT -based methods such as JPEG, the proposed approach achie v es higher com- pression g ains while maintaining competiti v e performance in PSNR, MSS IM, and VIF , conrming that quin- cunx sampling can eliminate redundant spatia l information without substantially compromising visual delity . Enhanced ima g e compr ession thr ough hybrid sta g g er ed downsampling and DCT ... (Benlabbes Haouari) Evaluation Warning : The document was created with Spire.PDF for Python.
68 ❒ ISSN: 2502-4752 These ndings are consistent with pre vious research on multi-resolution sampling and perceptual compression, while of fering a simpler and more computationally ef cient alternati v e to purely transform-based or neural netw ork–dri v en approaches. Future w ork could e xtend this frame w ork by adapting the do wnsampling rate ac- cording to local v ariance, inte grating perceptual quantization matrices, or incorporating lightweight learning- based renement modules to further enhance reconstruction quality . Additional studies e v aluating the method across di v erse datasets, illumination conditions, and image types w ould strengthen its rob ustness. In v estig a- tions in v olving color images, multispectral data, and real-time hardw are implementation are also important ne xt steps. Ov erall, this s tudy underscores that a carefully designed h ybrid spatial–transform pipeline remains a practical and ef cient strate gy for image compression, pro viding a solid foundation for future methodological de v elopments. Figure 3. Results - “Lena image” Figure 4. Results - “House image” Figure 5. Results - “Mandril image” Figure 6. Results - “Peppers image” 4. CONCLUSION This study introduces a no v el h ybrid image compression algorithm that inte grates do wnsampling with the DCT . The proposed approach achie v es an ef fecti v e balance between compression ef cienc y and image quality , outperform ing con v entional met ho ds such as JPEG and SPIHT+D WT across multipl e met rics, includ- ing PSNR, SSIM, VIF , b i t rate, and compression ratio. Be yond quanti tati v e performance, the method deli v ers Indonesian J Elec Eng & Comp Sci, V ol. 42, No. 1, April 2026: 62–70 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 ❒ 69 superior visual quality by reducing the blocking artif acts commonly observ ed in JPEG and by preserving ne te xtures more ef fecti v el y than SPIHT+D WT , which tends to obliterate or o v ersm o ot h image details. By com- bining do wnsampling with selecti v e retention of DCT coef cients, the algorithm attains high compression while maintaining strong perceptual delity . This mak es it particularly suitable for applications with limited storage or bandwidth, including remote sensing, embedded systems, video compression, and medical image preserv ation. REFERENCES [1] G. K. W allace, “The jpe g still picture compression standard, ” IEEE T ransactions on Consumer Electronics , v ol. 38, no. 1, pp. xviii-xxxi v , 1992, doi: 10.1109/30.125072. [2] C. K umar , “Hybrid optimization for secure and rob ust digital image w atermarking with dwt, dct and spiht, ” Multimedia T ools and Applications , v ol. 83, no. 11, pp. 31911–31932, 2024, doi: 10.1007/s11042-023-16903-8. [3] P . Y ang, C. Zhou, D. Shullani, L. Liu, D. and Baracchi, “ A comprehensi v e re vie w on le containers-based image and video foren- sics, ” Computers, Materials and Continua , v ol. 85, no. 2, pp. 2487–2526, 2025, doi: 10.32604/cmc.2025.069129. [4] H. Kim, A. No, and H.-J. Lee, “SPIHT algorithm with adapti v e selection of compression ratio depending on D WT coef cients, ” IEEE T ransactions on Multimedia , v ol. 20, no. 12, pp. 3200–3211, 2018, doi: 10.1109/TMM.2018.2832604. [5] S. Nirmalraj, “SPIHT : A set partitioning in hierarchical trees algorithm for image compression, ” Contemporary Engineering Sci- ences , v ol. 8, no. 6, pp. 263–270, 2015. [6] A. F ajar , R. Sarno, and C. F atichah, “ Ev aluating the impact of do wnsampling on 3D MRI images se gmentation results based on similarit y metrics, ” Indonesian Journal of Electrical Engineering and Computer Science , v ol. 35, no.3, 1590-1600, 2024, doi: 10.11591/ijeecs.v35.i3.pp1590-1600. [7] X. W u and N. Memon, “Conte xt-based, adapti v e, lossless image coding, ” IEEE T ransactions on Communications , v ol. 45, no. 4, pp. 437–444, 1997, doi: 10.1109/26.585919. [8] D. S. T aubman and M. W . Marcellin, “Jpe g2000: Standard for interacti v e imaging, ” Proceedings of the IEEE , v ol. 90, no. 8, pp. 1336–1357, 2002, doi: 10.1109/JPR OC.2002.800725. [9] K.-O. Cheng, N.-F . La w , and W .-C. Siu, “F ast e xtraction of w a v elet-based features from jpe g images for joint retrie v al with jpe g2000 images, ” P attern recognition , v ol. 43, no. 10, pp. 3314–3323, 2010, doi: 10.1016/j.patcog.2010.05.012. [10] M. V amsikrishna et al., “Re gion based lossle ss compression for digital images using entrop y coding, ” Indonesian J ournal of Elec- trical Engineering and Computer Science , v ol. 38, no. 3, pp. 1870-1879, 2025. [11] S. F adl, Q. Han, and Q. Li, “Surv eillance vide o authentication using uni v ersal image quality inde x of temporal a v erage, ” in Digital F orensics and W atermarking: 17th International W orkshop, IWD W 2018 , Jeju Island, K orea, Springer , 2019, pp. 337–350. [12] B. Haouari, “3d medical image compression using the quincunx w a v elet coupled with spiht, ” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS) , v ol. 18, no. 2, pp. 821–828, 2020, doi: 10.11591/ijeecs.v18.i2.pp821-828. [13] H. R. Sheikh and A. C. Bo vik, “Image information and visual quality , ” IEEE T ransactions on image processing , v ol. 15, no. 2, pp. 430–444, 2006, doi: 10.1109/TIP .2005.859378. [14] K. Seshadrinathan, H. R. Sheikh, A. C. Bo vik, and Z. W ang, “Structural and information theoretic approaches to ima ge quality assessment, ” in Multi-Sensor Image Fusion and Its Applications , CRC Press, 2018, pp. 473–500. [15] Benl abbes, Haouari. “Study of transmiss ion system for compressed and encrypted image, ” Multimedia System s , v ol. 27, no. 3, pp. 471-482, 2021, doi: 10.1007/s00530-020-00737-7. [16] D. R. I. M. Setiadi, “PSNR vs SSIM: imperceptibility quality assessment for image ste g anograph y , ” Multimedia T ools and Appli- cations , v ol. 80, no. 6, pp. 8423-8444, 2021, doi: 10.1007/s11042-020-10035-z. [17] Y . A. Najjar , “Compar ati v e analysis of image quality assessment metrics: MSE, PSNR, SSIM and FSIM, ” International Journal of Science and Research (IJSR) , v ol. 13, no. 3, pp. 110-114, 2024, doi: 10.21275/SR24302013533. [18] I. Bakuro v et al., “Structural similarity inde x (SSIM) re visited: A data-dri v en approach, ” Expert Systems with Applications , v ol. 189, p. 116087, 2022. [19] M. Martini, “ A simple relationship between SSIM and PSNR for DCT -based com pressed images and video: SSIM as content-a w are PSNR, ” 2023 IEEE 25th International W orkshop on Multimedia Signal Processing (MMSP) , IEEE, 2023. [20] J. W u, et al., “Reduced-reference image quality assessment with visual information delity , ” IEEE T ransactions on Multimedia , v ol. 15, no. 7, pp. 1700-1705, 2013, doi: 10.1109/TMM.2013.2266093. [21] C. Peng, M. W u, and K. Liu, “Multiple le v els perceptual noise back ed visual information delity for picture quality assessment, ” 2022 International Symposium on Intelligent Signal Processing and Communication Systems (ISP A CS) , IEEE, 2022. [22] Z. Jia et al., “Generati v e latent coding for ultra-lo w bitrate image compression, ” Proceedings of the IEEE/CVF Conference on Computer V ision and P attern Recognition , 2024. [23] T . Bordin and T . Mauge y . “Linearly transformed color guide for lo w-bitrat e dif fusion based image compression, ” IEEE T ransactions on Image Processing , 2024. [24] D. Si v araman, J. Jebanazer , and B. Balasubramanian. “Discriminati v e analysis of w a v elets for ef cient medical image compression. ” Indonesian Journal of Electrical Engineering and Computer Science , v ol. 30, no.1, pp. 510-517, 2023. [25] Z. Zhao et al., “Compression ratio allocation for probabilistic semantic communication with RSMA, ” IEEE T ransactions on Com- munications , 2025. Enhanced ima g e compr ession thr ough hybrid sta g g er ed downsampling and DCT ... (Benlabbes Haouari) Evaluation Warning : The document was created with Spire.PDF for Python.
70 ❒ ISSN: 2502-4752 BIOGRAPHIES OF A UTHORS Benlab bes Haouari is a researcher and lecturer in the eld of computer science and information technology . He is af liated with the Department of Exact Sciences at the Higher Nor - mal School of Bechar , Algeria. His research interests include cloud computing, virtual machine scheduling, distrib uted system s, wireless sensor netw orks, and image processing. He has contrib uted to se v eral publications related to cloud infrastructure optimization and multimedi a sensor netw orks, including studies on vi rtual machine load balancing and image compress ion techniques in wireless multimedia sensor netw orks. He can be contacted at email: hbenlabbes@yahoo.fr . Khair Y ounes recei v ed his de gree in computer science from the Uni v ersity of T ahri Mohamed Bechar , Algeria. He is currently af liated with the F aculty of Exact Sciences at the same uni v ersity . His research interests include cloud computing, virtualization technologies, resource man- agement in cloud en vironments, and virtual machine scheduling. His recent w orks focus on load balancing and ener gy-ef cient virtual machine migration in cloud computing infrastructures. He can be contacted at email: ynss.khair@gmail.com. Beladgham Mohammed recei v ed his de gree in computer science from the Uni v ersity of T ahri Mohamed Bechar , Algeria, where he is currently a f aculty member in the Department of Computer Science. His research interests include image processing, multimedia systems, wireless multimedia sensor netw orks, cloud computing, and data compression techniques. He has contrib uted to se v eral scientic publications focusing on image compression, multimedia transm ission, and op- timization techniques for distrib uted and cloud-based systems. He can be contacted at email: be- ladgham.tlm@gmail.com. El Hendi Hichem recei v ed his de gree in computer science from the Uni v ersit y of T ahri Mohamed Bechar , Algeria, where he is currently af liated with the Department of Computer Sci- ence. His research interests include cloud computing, distrib uted systems, wireless multimedia sen- sor netw orks, and information systems. His research acti vities focus on resource management, data transmission optimization, and multimedi a applications in distrib uted computing en vironments. He can be contacted at email: elhendi.hichem@uni v-bechar .dz. Indonesian J Elec Eng & Comp Sci, V ol. 42, No. 1, April 2026: 62–70 Evaluation Warning : The document was created with Spire.PDF for Python.