A New Perspective on Improving the Lossless Compression Efficiency for Initially Acquired Images

Abstract

A new lossless compression scheme of compressing the initially-acquired continuous-intensity images with a lossy compression algorithm to obtain higher compression efficiency is proposed. Even if a lossy algorithm is employed, for decoded original images, there is no loss of data in the same sense as the conventional lossless scheme. To realize the new idea, the compression efficiency of the existing lossy subband compression algorithm is improved at high bitrates. For the entropy coding part, a run-length based, symbol-grouping entropy coding method is introduced. For the quantization part, the entropy-constrained scalar quantization is implemented using a novel and simple thresholding method. Coding results show that bit savings of the proposed lossless scheme, which employs a lossy algorithm, over the conventional lossless scheme achieve a maximum of 27.2% and an average of 11.4% in our test.INDEX TERMS Lossless image compression, lossy subband image compression, compression efficiency at high bitrates, Golomb coding, run-length based entropy coding, entropy-constrained scalar quantization.

Reversible Integer-to-Integer Wavelet Filter Design for Lossless Image Compression

“…Coding tests on real images were performed to test the new reversible ITI 17/11 filters. An efficient entropy coding algorithm that uses the symbol grouping method and the modified Golomb-Rice coding [23] was used to code the integer coefficients decomposed from the reversible ITI transforms.…”

“…Nevertheless, the combination of the new 17/11 filters with the more efficient entropy coding still has the highest efficiency and achieves bit rate reductions of about 3.1% and 3.8% over JPEG-LS and JPEG2000 lossless mode respectively. As described in [23], the entropy coding was designed to more efficiently handle the noise in decomposed coefficients.…”

Modified Hilbert Curve for Rectangles and Cuboids and Its Application in Entropy Coding for Image and Video Compression

“…These binary arithmetic coding algorithms are practical algorithms because the coding of a multiple symbol source can always be converted to coding of a sequence of binary symbol sources. For example, in image compression, the bit-plane coding method [ 6 ] and the symbol grouping coding method [ 7 , 8 ] eventually convert the quantized coefficients to binaries to code.…”

“…For example, to code a binary source with the probabilities for the symbol “0” and for the symbol “1”, arithmetic coding needs to code 999 “0”s on average before it codes a “1”. On the other hand, run-length-based entropy coding [ 7 , 8 ] is much more computationally efficient for low-entropy coding situations, as it does not need to code the “0”s one by one. Note, low-entropy sources are very common in compression.…”

“…For non-stationary binary sources, the binary arithmetic coding algorithms use probability estimators to adapt to the probability variations; whereas the run-length-based entropy coding uses the symbol grouping method to handle non-stationary binary sources [ 7 ]. For coding 2-dimensional (and higher-dimensional) subband coefficient arrays, binary arithmetic coding can estimate the probabilities from the coded adjacent coefficients in different spatial directions (context modeling).…”