vendredi 1 juin 2012


New Release 0.1.6

Link: Updated 02/17/2019  windows binaries / source code on GitHub

New quality settings (exp): -l19->-l1,norm, -h1->-h3

12/17/2018: v0.1.6 very high compression added. This version is still fast with a good neatness. 

10/09/2018: I have developped NHW Neatness Metrics to measure neatness of images.

          An Image Comparison of NHW and x265     Sample 512x512 test images

                      Here are the 3 test images compressed with this new version 0.1.6:
                                         
Lighthouse1.nhw  30402octet
                                        
Birds5.nhw  44227octet
                                        
Butterflies2.nhw  41871octet



Remark: The entropy coding schemes are not totally optimal for now, I have the fast ideas to improve them, and so we can save 2.5KB per .nhw compressed file in average.

Nov2012 version now uses the YCbCr color space, as used in JPEG for example.YCbCr is more accurate.This new version also has a better color balance, 2 precision improvements and some bug fixes.

Next steps are: coding the other (lower and higher) quality settings, lower quality setting is rather complex...

Do not hesitate to leave any comment on this new version.

Many thanks,
Raphael
nhwcodec@gmail.com

jeudi 8 mars 2012

Hello,

Here is the demo page of the NHW Project.

The aim of the project is to create an advanced and state-of-the-art wavelet-based open-source codec that can compete with reference ones, such as AOM AV1, HEVC, Google VP9, Xiph Daala.It introduces some innovations like, for example, a new fast wavelet transform, a new multistage residual coding, new compression & entropy coding schemes.The NHW codec is also royalty-free, and the NHW Project does not own patents.

A main feature of the NHW codec is that it is optmized for speed.The codec is actually a lot faster to encode/decode than the codecs above.It is also optimized for visual quality.The codec has a good amount of neatness and gracefully retains or improves image quality/sharpness, whereas the reference schemes tend to decrease image neatness/sharpness but keep a good amount of details and have a good precision in return.


To test it:

The current NHW version is a very first and experimental one.It allows to compress for the moment a 512x512 bitmap color image for quick test and demo purpose (will have to adapt it to any size and dimensions of image).
The wavelet transform is performed on the whole image, this is a difference with the DCT block-based transform that is used in reference codecs, such as JPEG for image.

Remark: If you classically want to compress a (larger) JPEG image, you can for example use Paint software under Windows to select the 512x512 part that interests you of the JPEG image (if larger) and also to save it as a bitmap color image.

Many thanks,
Raphael
nhwcodec at gmail.com
                   Here are some very few results on some images (bitmap color 512x512):



Original: Lighthouse1 - bitmap color 786486octet

Compressed .nhw file: Lighthouse1.nhw 45934octet


Original: Birds5 - bitmap color 786486octet

Compressed .nhw file: Birds5.nhw 73101octet


Original: Butterflies2 - bitmap color 786486octet

Compressed .nhw file: Butterflies2.nhw 65968octet





Here are very few sample images, do not hesitate to test it on other images (bitmap color 512x512 for the moment) by downloading the win32 binaries linked above.


Remark: The entropy coding schemes are not totally optimal for the moment, and normally we can save or win 2KB-2.5KB in average on the compressed .nhw files.-These improvements will have to be done...-



Hello,

My name is Raphael Canut, this codec is my own work, (I have a Master's degree in Electrical Engineering, '2006).

NHW Codec would originally stand for New Hybrid Wavelet Codec... The original idea was to apply the classic compression techniques with intra residual coding on the first order wavelet DC image, and not on the whole image, and then use some properties, like neatness, with the higher order reconstruction wavelet filterbank...

Do not hesitate to send any comment, advice,...

Many thanks,
Raphael