Ref.: Ms. No. PAAA-D-10-00062

Rotation Invariant Features for Color Texture Classification and Retrieval under Varying Illumination

Pattern Analysis and Applications

Reviewer1-

Reviewer #1: All the major points seem to have been addressed - and in particular the authors have made it clear that it is image not surface rotation that they are dealing with.

- No correction required.

Reviewer 2- :

Sl.No / Comments / Modification / Page NOs
1 / In the Abstract, 'Rotation' should not be capitalised and some slight grammatical changes are required (e.g., 'The quaternion representation of color is shown to be effective, which treats color ...")
/ Correction done.
Grammatically corrected. / Page 1
2 / There are still several grammatical and stylistic problems:
Spacing page 1, line 19, col 2.
Grammar page 2, line 6, col 1.
/ Correction done. / Page 1 &2
3 / In equation (2) the norm of L is identically one. So is L the "normalised" illumination vector? Surely the image intensity (i.e., the amount of light energy reflected from a Lambertian surface) depends on the intensity of the illuminant and the angles to that surface. Section 2.1 needs reworking to describe the correct physics of the illumination process. It clearly doesn't apply in the case of a highly specular surface for example.
/ Yes. It is normalized illumination vector. / Page 3
4 / Section 2.2 should probably cite:
Koen E.A. van de Sande, Theo Gevers, Cees G.M. Snoek, "Evaluating Color Descriptors for Object and Scene Recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1582-1596, September, 2010 / Reference is included.(16) / Page 2
5 / In Section 2.3 do the authors mean quaternion principal component analysis when they refer to "quaternion principle analysis"?
/ Yes. It is Principal component analysis. It is corrected. / Page 4 Column 2 , paragraph 2.
6 / Is the LMS color space well known in the general imaging community? Please reference or define.
/ Yes.
Ref: Kanan, C. & Cottrell, G. W. “Robust Classification of Objects, Faces, and Flowers Using Natural Image Statistics”, In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010),pp.2472-2479, 2010. / -
7 / Section 2.4: The explanation of quaternion rotation is not clear. Why is "(0, I_q)" a 3D vector? / 3 color channels form a 3D quaternion vector. Each dimension corresponds to a color channel. / Page 5
8 / Would the authors please disclose the "15 selected textures" or were they randomly selected? / Yes they are randomly selected to cover different classes of materials. / Page 8

Reviewer #3: All major points of concern in my initial review have been addressed sufficiently in this revised version. Hence, I recommend to accept this manuscript.

-No correction required.