A regularized constrained iterative restoration algorithm for restoring color-quantized images

Artificial Intelligence

“…We see, through intensive experiments, that 0 -AP significantly outperforms methods in [24,129,130], and also 1 -AP.…”

“…Regularised Constrained Iterative Restoration. In [129] it is described a method to remove quantisation artifacts, called Regularised Constrained Iterative Restoration (RCIR). The method is described applied to vector quantisation.…”

“…0 -AP vs. existing methods Now we compare the 0 -AP performance vs. that of the methods previously described: RCIR [129], CD [130], and DT+OP [24]. Table 6.1 shows the performance average (obtaining the MSE for the average) of each method in our test set (see Appendix A) for all the range of possible quantisation bits in 8-bits images.…”

“…Up to our knowledge, the first work was [128], which used de-quantising as a previous step to edge detection. Recently, other methods have been published, based on iterating between some filtering operation and the correction of the difference with the original [129,130]. But this type of strategies, though they result in efficient algorithms, are too simple to provide satisfactory results.…”