|Signal Processing: Image Communication
|Volume 37, 2015
|Stephan A. Rein; Frank H.P. Fitzek; Clemens Gühmann; Thomas Sikora
|Evaluation of the wavelet image two-line coder: A low complexity scheme for image compression
|This paper introduces the wavelet image two-line (Wi2l) coding algorithm for low complexity compression of images. The algorithm recursively encodes an image backwards reading only two lines of a wavelet subband, which are read in blocks of 512 bytes from flash memory. It thus only requires very little memory, i.e., a memory array for two wavelet subband lines, an array to store intermediate tree level data, and an array for writing binary data. A picture of 256×256 pixels would require 1152 bytes of memory. Computation time for the coding is derived analytically and measured on a real system. The times on a low-cost microcontroller for 256×256 grayscale pictures are measured as 0.25–0.6 s for encoding and 0.22–0.77 s for decoding. The algorithm can thus realize a low complexity system for compression of images when combined with a customized scheme for the wavelet transform; low complexity here refers to low memory, minimum write access to flash memory, usage of integer operations only, and low conceptual complexity (ease of implementation). As demonstrated in this paper, a compression performance similar to JPEG 2000 and the more recent Google WebP picture compression is achieved. The compression system uses flash memory (SD or MMC card) and a small camera sensor thus building an image communication system. It is also suitable for mobile devices or satellite communication. The underlying C source code is made publicly available.
|compression, Wavelet image two-line coder (Wi2l); Set-partitioning in hierarchical trees (SPIHT); Backward coding of wavelet trees (BCWT); Fractional wavelet filter; Fixed-point wavelet transform; Camera sensor node