Quantitative Evaluation of Lossy JPEG Compression Executed in a Digital Camera

概要

論文の詳細を見る
The recent increase in number of pixels of images acquired by a digital camera encourages one to utilize it for image measurement. A large number of pixels of the image bring difficulties in dealing with it. Therefore, lossy image compression is going to be required. Almost all digital cameras have the built-in function of lossy JPEG compression. Since there are few reports on evaluation of image matching accuracy using a pair of lossy compressed images acquired by a digital camera, we decided to investigate effects of lossy JPEG compression executed in a digital camera on the accuracy of image matching. This paper reports the preparatory experiment conducted in order to evaluate lossy JPEG compression quantitatively by using 54 diverse images. Two compression parameter sets of a downsampling ratio, a quantization table, and a Huffman code table utilized in Canon EOS 20D were investigated. The experiment results demonstrate that the compression ratio is correlated to grey level difference vector measures. Furthermore, the experiment results suggest that the degradation of quality of images reconstructed from images compressed in a digital camera is so small that it may have no effect on the accuracy of image matching.
社団法人日本写真測量学会の論文