Image compression is a form of data compression frequently used in digital imaging applications to improve the efficiency of transmission and storage. In order to have a manageable file size, the data compression is lossy, i.e. image information is lost. Lossy compression algorithms such as Baseline JPEG use the insensitivity of the human eye to high frequency information as a basis for the compression. Discrete cosine transforms (DCT) are performed on the data followed by variable quantization. And the image blocking has no exact regulations for the spatial variation. Using this technique, information is lost and the decompressed image is a distorted version of the original. It is known that simple repeated compressions do not influence image quality, which is deter mined by the initial compression. Considering practical use, this study extends JPEG operation to the repeated DCT block rearrangement operation of the JPEG baseline scheme. In particular image placement and translation are discussed. Objectively and subjectively analytical results are presented. It is now shown that these operations affect total image quality. The error of the JPEG image quality tends to increase with the number of repetitive DCT rearrangement operations, and this increase is dependent on the image frequencies (error is concentrated in the image's high frequency regions). Also, the artifacts become worse with lower quality levels of compression, or higher compression ratios. If the JPEG scheme could be modified to recognize the translations by compensating for the crop or translation, the artifacts due to this translation would be eliminated. Consequently, this research will attempt to investigate the error resulting from such translations.

Library of Congress Subject Headings

Image processing--Digital techniques--Quality control; Image compression; Computer algorithms; Image processing--Digital techniques--Data processing

Publication Date


Document Type


Department, Program, or Center

School of Print Media (CIAS)


Romano, Frank


Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TA1637 .T56 2000


RIT – Main Campus