Examples of Our Work

Interactive visual information processing, interactive image retrieval, segmentation .... (link ...)

Example-based Image Processing  ...  (link ...)

Unifying histogram-based image content descriptor for content-based image retrieval - the CPAM and AI framework  ...

Integrated framework for image coding, indexing, and content-based image retrieval ... (link ...)

Visual guided browsing and navigation for fast image retrieval ... (link ...)

Fast comprehensive tone mapping for high dynamic range image visualization ... (link ...)

Colorizing black and white photos ... (link ....)

Making beautiful pictures,  computational photography ... (link ....)

Machine learning, pattern recognition ... (link ...)

Image data mining, image and feature co-clustering .... (link ...)

Unifying Histogram-based Image Descriptors for Content-based Image Retrieval

Histogram-based image content descriptors are extensively used in content-based image retrieval, from color histogram to color correlogram. In this work, we have developed a method that can be regarded as unifying histogram based image content descriptors. Our approach is to first generate achromatic and chromatic image patches (spatial patterns) that are statistically most representative of natural images, and then to collect the statistics of these patterns appear in the image. Setting our patch size to 1x1 pixel, our descriptor becomes the color histogram descriptor. Setting the patch shape/size to two pixels D-distance apart, our descriptor becomes the correlogram descriptor, etc. ...

G. Qiu, "Appearance indexing", Proceedings of  ICASSP'03, IEEE International Conference on Acoustics,Speech, and Signal Processing, vol. III, pp. 597 - 600, April 6 – 10, 2003, Hong Kong  (PDF) 
G Qiu, "Indexing chromatic and achromatic patterns for content-based colour image retrieval", Pattern Recognition, vol. 35, pp. 1675 – 1686, August, 2002   (PDF)
G. Qiu, "Image indexing using a coloured pattern appearance model", Proc. Storage and Retrieval for Media Databases'2001, 21-26 January 2001, San Jose, CA, USA

The color and pattern separable model (+VQ) used to generate representative patches

Some examples of statistically representative patches. Interestingly, even though these paterns are "learned" from natural images unsupervised, some patches resemble edge detector, some resemble corner detector etc. ...