Ubuntu 16 OpenCV 2.4

Problems during compilation:

relocation R_X86_64_32 against `ff_a64_muxer' can not be used when making a shared object

Solution:
recompile ffmpeg-3.4.5 with(2018-11-29):
./configure --enable-nonfree --enable-pic --enable-shared

Download ffmpeg and build 3.4.5, 3.2.12 or 4.0 and try to compile opencv doesn't help, then

Final solution: disable ffmpeg
$cmake  -D WITH_CUDA=OFF -D WITH_FFMPEG=0 ..
$make

That compile opencv 2.4.x and 3.4.x



References:
[1] Error sys/videoio.h not found problem http://yutopapa.hatenadiary.com/entry/2017/06/08/173149

3D Models & tools

1) Open .obj 3D format

view3dscene
$sudo apt-get install view3dscene

meshlab - System for processing and editing triangular meshes
$sudo apt-get install meshlab

g3dviewer
$sudo apt-get install  g3dviewer

Other tools

1. glc_player which is said to read-and-show '.3ds', '.obj', '.stl', '.off', '.3dxml', and Collada ('.dae') files
2. g3dviewer which is said to read-and-show '.3ds', '.lwo', '.obj', '.dxf', '.md2', '.md3', '.wrl', '.vrml', '.dae' (COLLADA), '.ase' (ASCII Scene Exporter), '.ac' (AC3D)
3. ivview which reads-and-shows '.iv' and VRML1 files
4. paraview which reads-and-shows '.ply' and '.vt*' files
5. varicad-view which reads-and-shows '.dwg' (2D), '.dxf' (2D only?), '.igs' (maybe?), '.stp' (3D) files 
6. wings3d - Nendo-inspired 3D polygon mesh modeller (legacy)
7. gmsh - Three-dimensional finite element mesh generator
8. libadmesh-dev - Tool for processing triangulated solid meshes.
9. libgmsh-dev - Three-dimensional finite element mesh generator.
10. libmadlib-dev - mesh adaptation library
11. libnglib-dev - Automatic 3d tetrahedral mesh generator development files
12. libscotch-dev - programs and libraries for graph, mesh and hypergraph partitioning
13. netgen - Automatic 3d tetrahedral mesh generator
14. libtriangle-dev - High-quality 2-D mesh generator development files

Resources
 [1] mview http://mview.sourceforge.net/

Models
  [1] .max, .obj, blender, etc. www.turbosquid.com
  [1] http://tf3dm.com/

Lectures Neural Networks

Colour and Texture, but fail for shape.

Generative Adversarial Networks
https://arxiv.org/pdf/1406.2661.pdf

Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks https://junyanz.github.io/CycleGAN/

Journals for Bio Informatics

https://scfbm.biomedcentral.com/track/pdf/10.1186/1751-0473-3-6?site=scfbm.biomedcentral.com

springer computer vision http://www.springer.com/computer/image+processing/journal/11263  11 8.2

IEEE http://signalprocessingsociety.org/publications-resources/ieee-transactions-image-processing 44 4.3
ELSEVIER Pattern recognition https://www.journals.elsevier.com/pattern-recognition/ 47 4.5
ELSEVIER Medical image analysis https://www.journals.elsevier.com/medical-image-analysis/ 56 4.1
IEEE Medical images https://ieee-tmi.org/ 68 3.9
ELSEVIER https://www.journals.elsevier.com/computer-vision-and-image-understanding/ 112 3.2
ELSEVIER https://www.journals.elsevier.com/image-and-vision-computing/ 165 2.6
ELSEVIER https://www.journals.elsevier.com/computer-vision-and-image-understanding/ 189 2.4



http://www.guide2research.com/journals/computer-vision

PCA Feature extraction

References:
[1] PCA http://www.visiondummy.com/2014/05/feature-extraction-using-pca/
[2] Reducción de dimensinalidad usando PCA   https://www.coursera.org/learn/clasificacion-imagenes/lecture/PaTVm/reduccion-de-descriptores-pca


[3] Opencv code with explanation for dimentional reduction
    https://stackoverflow.com/questions/27733002/how-to-use-pca-to-reduce-dimension
[4] Distances http://wwwae.ciemat.es/~cardenas/docs/lessons/MedidasdeDistancia.pdf

Distances

References
[1] http://wwwae.ciemat.es/~cardenas/docs/lessons/MedidasdeDistancia.pdf

Bag of Features and Texture

Notes

What is the difference between SIFT and Dense SIFT

*SIFT consists of both detection and description while dense sift only uses the descriptor in densely sampled locations [1].

*SIFT identifies interest points using Difference of Gaussian Filtering (DoG) before using Histogram of Oriented Gradients (HOG) to describe these interest points, however Dense-SIFT does not identify interest points, it simply divides the image into overlapping cells before using HOG to describe them. since they both use HOG they both produce 128 dimensional feature vectors [1].

*SIFT is typically computed at interest points. Dense SIFT is computed at every pixel, or every kth pixel. HOG is computed for a rectangular cell array where each cell is usually 8x8 pixels. Dense SIFT and HOG are similar in the sense that they both characterize edginess and orientation around pixels, but the computations are different. Jianxiong Xiao's 2x2 HOG is different than normal HOG. The truth is that once you know how these kinds if features work you can get fancy and histogram them differently, change normalization terms, etc and create your own variant. I spoke with Prof Xiao many times about this when we ovarlapped at MIT [2].

*Firstly, Difference of Gaussians (DoG) can be used for estimating Laplacian of Gaussians (LoG), which are useful for finding edges and blobs. DoG is computationally faster so it is used. Overall, the way in which LoG is used for SIFT and HOG is the fundamental difference between these two feature descriptors. Dense SIFT is exactly as it sounds, SIFT computed densely for every pixel in the image and it helps in image registration, pose estimation, object recognition, etc [2].

Resources

1) BoF imlmentation using SURF, IHOG http://www.cvc.uab.cat/~aldavert/plor/software.html

2) Texture video https://www.youtube.com/watch?v=LQBKIi-Xtbc

3) Textons http://webpages.uncc.edu/~yjaved/publications.html


4) Bag of Visual Words implementation (Functional) http://www.codeproject.com/Articles/619039/Bag-of-Features-Descriptor-on-SIFT-Features-with-O

5) Gabor filters histogram, explanation
    http://stackoverflow.com/questions/20608458/gabor-feature-extraction   
6) Filter Banks, Matlab Source Code
    http://www.robots.ox.ac.uk/~vgg/research/texclass/filters.html

7) Texture classification using textons  http://courses.media.mit.edu/2008fall/mas622j/Projects/NickLoomis/


References
 [1] https://www.researchgate.net/post/What_is_the_difference_between_SIFT_and_Dense_SIFT
 [2] https://www.quora.com/Computer-Vision-Is-there-a-difference-if-any-between-dense-SIFT-and-HOG

Fiducial Land Mark (Face Keypoints)

Facial Keypoints
[1] https://github.com/sunsided/facial-keypoints
[2] kaggle https://www.kaggle.com/c/facial-keypoints-detection/data


Resources:
1) Source Code Matlab/C++
http://cmp.felk.cvut.cz/~uricamic/flandmark/

Different kind of Image Medical Datasets

References:
[1] Neuromorph https://datamed.org/repository_list.php
[2] http://neuroangio.org/anatomy-and-variants/anterior-cerebral-artery/

Common datasets for image processing and computer vision

Resources:

[1] Hela https://ome.grc.nia.nih.gov/iicbu2008/hela/index.html

Funny Apps

Photomath
https://photomath.net/en/

Lectures Image Processing

Resources:

1) Notes http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/MORSE/
2) Wavelets teory (pt) http://rmu.sbm.org.br/Conteudo/n33/n33_Artigo02.pdf

Image Sharpen

References
[1] http://www.idlcoyote.com/ip_tips/sharpen.html

Microscopy Software

CISMM
http://cismm.cs.unc.edu/downloads/

Image Processing tools

[1] BioImage Suite
http://download.bioimagesuite.yale.edu/download/Install_2.5.html