Streamlit normalization

[ThijsVroegh]

This PR adds a new page to the dashboard for image alignment, centered on removing differences in rotation, scale, translation, and shearing. After normalization on brightness, sharpness, contrast and colour, alignment will take place and changes in rotation, scale, translation, and shearing will be displayed. Covered are:

1. Based on a selected base-image, all other images will be aligned to this base image accordingly
2. Multiple file upload is possible
3. Option to select from several alternative alignment procedures
4. Clear visual inspection and comparison by having an image viewer build in to scroll between all aligned images
5. In addition to saving into buffer, also provide the option to save the aligned images locally

The aligned images can be saved and used for further 1) outline contour extraction and comparison and 2) structural comparison of outlines form the images. These are handled in spearate streamlit pages for clarity purposes

TODO:

• [X] Option to save aligned images
• [x] Save into buffer to be passed on to next page/ analysis
• [x] Add as page to existing dashboard (with similar layout)
• [X] Add equalization steps, in terms of brightness, color, sharpness, contrast
• [X] Add to feature-based alignment the option to choose ORB, SIFT, or SURF (currently, ORB is used)
• [X] Add to Fourier Mellin Transform the option to choose dont normalize, normalize, or phase
• [X] Make base image and aligned image equal in size in the viewing part when they are alongside each other for optimal visual comparison
• [x] Homography decomposition: from the homography or similar type of estimated matrix, depending on which procedure is used, extract information on difference of several image parameters (e.g., degree of rotation). This information will be put below the aligned image. So, given a selected base image, we get for all other aligned images an overview of differencnes in:
• [x] rotation
• [x] scale
• [x] translation
• [x] shearing
• [X] Add several alignment procedures to choose from:
• [X] Feature based registration (Orb)
• [X] Enhanced Correlation Coefficient (ECC) maximization
• [X] FFT phase correlation
• [X] Fourier-Mellin
• [x] Add animation of base and aligned image for further visual inspection
• [x] Add slider to directly compare base and aligned image using appropriate tool (https://github.com/fcakyon/streamlit-image-comparison)
  • [x] Add values on brightness, sharpness and contrast (after having clicked to normalize these)

[stefsmeets]

I refactored the code a bit in line with our discussions this morning.

Todo

• [x] Do not depend on config widget
• [x] Check if opencv can be replaced by scikit image -> no longer necessary