similar to: NHW Codec - next lower quality settings

Hello, As I said in my previous post, I would like now to accelerate to NHW Codec development.I started in January 2007, more than 10 years ago, and I would like now to see it finished after all the time I gave to it (especially the 3 first years).For now I can only give 2-3 hours in the week-end to the codec development, and furthermore I don't have all the skills to improve it, notably in

Hello, I have improved a little the NHW codec.I have improved the dequantization in the wavelet HH band, and little improved the -l2 (and lower) quality setting.I have also added a -l3 lower quality setting (-15Ko, exp), I have increased the quantization for this other setting, and I start to remove higher wavelet coefficients, but this is not optimal... The artifacts start to be noticeable for

Hello, I was testing quickly some other lower quality settings -l3, -l4 because high compression is what really lacks to the NHW codec. In fact I can remove -5Ko of residual coding on Y,U,V comps for -l3 and I can remove -10Ko always of residual coding for -l4, and -l3, -l4 settings will be really better than the current ones implemented. I can also increase the wavelet coefficients threshold

Hello, I have finally added 2 lower quality settings for the NHW codec: -l1 (-5Ko) and -l2 (-10Ko).I use a quantization of 0.935 and 0.88 (kind of quantization), and I decrease residual coding on the first order wavelet image. I have updated the demo page: http://nhwcodec.blogspot.com/ . These 2 lower quality settings are still experimental.If you could find time, would be interested in any

Hello, In my last post, I seemed to say that there were no directional wavelet transforms that would give improvement to the NHW codec, but are directional wavelets an active research field today? I have also seen some recent patents on advanced and improved SPIHT, and I am a little surprised that there is no "up-to-date" state-of-the-art wavelet codec.Rududu in April 2008 was the last

Hello, Just a quick message, I have improved the -l3 lower quality setting.I have increased the quantization to 0.8 and decrease a little the denoising, and so to save 1Ko in average, I have added a 8x8 block variance analysis.Blocks that have a low variance are averaged to save bits.The -l3 lower quality setting is a little better now. For the -h3 higher quality setting, yes for now I do not

Hello, I am still progressing with very high compression and I have corrected a processing, so now very high compression (-l10 to -l13 quality settings) has really better precision and so better quality. I have also released a first draft of a new -l14 very high compression quality setting. More at: http://nhwcodec.blogspot.com/ Just a remark, the entropy coding schemes start to be less

Hello, Definitely not any (constructive) comments on the NHW Image codec? According to my tests, it can compare good with reference codecs like x264 intra and WebP. Yes, the codec is still experimental and is optimized for visual (subjective) evaluation, not PSNR.Its main feature is again that it would have a little more neatness/sharpness and a little less precision (residual coding is applied

Hello all, Just a quick message for those interested in the NHW Project.I realize that I have too little time to spend on the project currently and I feel that this year will be very important after 6-7 years for its success or not.I already had some few replies from some companies (and a major one), and they can not spend some time, spend some engineers on my work because there is no

Hello, Just a very quick message to let you know that I have added a higher quality setting (-h2, +10Ko) and I have added now residual coding on the 512x256 wavelet image for the higher settings.Precision is a little better.-I also still apply the post-processing function on the 512x256 wavelet image for now.- Here is the link for the binaries (as I can not link them in my blog for now):

Hello, I would like to discuss a little about the NHW codec and its state after nearly 5 years on the Xiph forums. I also know that if I would like to discuss I'd better go to the great Doom9 and Encode.ru forums but for now I find that my codec is too experimental for these professional forums: - the codec is only for fixed 512x512 size, just a test tool. - there is only mid compression -

I don't mind at all. I imagine there are more people like me out there who have been following your work, but haven't said anything. I agree that a series of pages explaining how things work may lead to more contributions. At the very least it would lead to more people being better educated on how wavelets work in image compression. But, this is your project, if you don't have the

Hello, I have re-tested -l4 high compression setting and it's clear that it lacks of precision on degraded, rather blurred images. So I don't know if it is a good idea to base the other lower quality settings (-l5,-l6,...) on -l4 setting.I have tested the NHW codec against x265, x264, Daala, WebP, Rududu, DLI and it's clear that at high compression these very good codecs have more

Hello, I forgot in my last reply that my DWT implementation can be speed up, for example I'm doing for now: for (;_X1<_E_;_X1++,_RES+=2) //dilatation { _RES[0]=_X1[0]<<3; _RES[1]=(_X1[1]+_X1[0])<<2; } then for (;_X2<_E_;_X2++,_RES+=2) //details { _RES[0]-=(_X2[1]+_X2[0])<<1; _RES[1]+=6*_X2[1]-_X2[2]-_X2[0];

Hello, So here is the compression scheme that I don't know if is covered by patent. I use it on the wavelet DC parts.It outputs 1 byte (8 bits) words.There is 5 modes: - 1(MSB) | 7 bits to store value (0->127 range) - 01 | 6 bits to store val[n+1]-val[n] | val[n]-val[n-1] if each in the range [-4,4] - 001 | 5 bits to store val[n+2]-val[n+1] | val[n+1]-val[n] | val[n]-val[n-1] if each in

Hello, No advance for now on the NHW codec, I am still working on a new quantization scheme to scale down the compression ratios but without luck... Any advice, idea is very welcome.I will also have to correctly code the wavelet filerbank and process on rows and columns and suppress the image transposition (this is for long overdue)... will win few cycles... there are other parts to speed up (in

Hello, I don't have advanced on the NHW codec, but now, as a lot of people, I am following the AOMedia AV1 codec.Just a (late) update of my previous post, there won't be problem in AOM AV1 with the CABAC scheme, as the codec will use Daala's multi-symbol entropy coder and/or ANS.That's great! I have even read on the wiki page, that AV1 will use Daala's PVQ as quantization

Hello, I have just tested new next lower quality settings (-l3 and -l4), I can diminish residual coding but if I increase wavelet coeffs threshold there is more aliasing.And even in the current mid quality settings there is some aliasing, this is the big drawback of the NHW codec. SPIHT algorithm is good for that as for example Rududu codec doesn't have aliasing at all at mid compression and

Hello all, It's me again.Just a quick message to tell that I finally do not agree that HEVC (don't have tested AOM AV1 for now) is better than the NHW codec at mid compression. Last night, I dowloaded 30 images of Star Wars Rogue One on the web and I made another comparison with x265 (HEVC) for mid compression (-l2 to h2 settings for the NHW codec). Visually, I prefer the results of NHW

Hello, I have improved quality for -l1->-l4 settings.I have properly set up the wavelet coeff threshold of the dead-zone according to the different wavelet subbands, I have then saved another 1Ko per .nhw files without noticeable difference and so I have raised again the quantization.The reconstructed images have now a better precision and still a good neatness. Update at: