Электроника

 
15 | 12 | 2018

It has been proved in the compression society that wavelets have strong potential for lossless and lossy image compression. Wavelet image coefficients are characterized by more narrow distribution of zero-order entropy in comparison with that of Discrete Cosine Transform (DCT), which is quite critical for entropy coding.

 

Any wavelet codec is free of such artifacts as blocking effect – the main disadvantage of JPEG standard. The use of consequent wavelet filtering at different space scales (pyramidal levels) is well suited for elimination of spatial redundancy of pixel information, especially in the case of high resolution images, contradictory to small block size of 8x8 pixels used in JPEG. Moreover, internal correlations of wavelet coefficients allow for state-of-the-art coefficient coding, one of which is the well known Shapiro’s Zero-Tree wavelet coefficient coding. For these reasons during several last years wavelets have attracted very strong attention and a new image compression standard based on wavelets, JPEG 2000 (or JPEG2K), is now considered for approval.

In spite of the fact that many wavelet based coding methods have been proposed the novel IDM wavelet codec has the best performance up to the date.

Example 1 shows the rate-distortion curves for standard test image “Lena” (resolution 512x512). The PSNR improvement is about 1 dB in comparison with JPEG 2000 and about 0.5 dB in comparison with the codec of Xiong based on trellis quantization. IDM wavelet codec is also 1 dB better than very well known SPIHT codec.

Example 2 demonstrates the power of IDM codec for losseless compression. The file size is the smallest among considered examples.

A free demonstration program with implementation of IDM wavelet image encoding algorithm is available. This program performs the conversion of RGB24 BMP file into IDM Wavelet format and vice versa.

Example 1. Rate-distortion curves for Lena test image (512 x 512 pixels) obtained with IDM and other popular lossy wavelet-based image compression methods:
Lena test image (resolution reduced)

 

Example 2. Lossless compression of test image

The photo was taken with the use of professional 6-Mpixel digital camera Kodak DCS 760 having resolution 2008 x 3032 pixels. The image resolution was further reduced to 1024 x 1408 pixels.

 

IDM compression algorithm is compared to WinZip and PNG-24.

Original size: 4,125 MB
Compressed by WinZip: 2,693 MB
Compressed by PNG-24: 1,728 MB
Compressed by IDM Wavelet: 1,279 MB

Digital camera Kodak DCS 760 (resolution reduced)