[WIP] refactor codebase with functor APIs

[johnnychen94]

I found filter specification greatly reduces the complexity of code structure and improves the style consistency when I played with ImageQualityIndexes and ImageNoise. Check my GSoC 2019 blog The principles of Images.jl: Part I as an example and illustration. Also, see Proposal: add abstract types ImageAlgorithm and ImageFilter for a unified API that enables extensibility without loss of API consistency.

To make this PR easy to review, the first eight commits serve as a start point, future commits are submitted in the form of PRs.

TODO:

• [x] move implementations into concrete functors, swap arguments, and refactor docstring (closes #23 )
  • [x] AdaptiveThreshold PR: https://github.com/zygmuntszpak/ImageBinarization.jl/pull/30
  • [x] Balanced PR: #32
  • [x] Entropy PR: #33
  • [x] Intermodes PR: #34
  • [x] MinimumError PR: #35
  • [x] MinimumIntermodes PR: #36
  • [x] Moments PR: #37
  • [x] Niblack PR: #42
  • [x] Otsu PR: #38
  • [x] Polysegment PR: #44
  • [x] Sauvola PR: #43
  • [x] UnimodalRosin PR: #39
  • [x] Yen PR: #40
• [x] adopt the API proposed in #41 (closes #41) PR: #45
• [x] update documentation and README (closes #25)

[codecov[bot]]

Codecov Report

Merging #29 into master will decrease coverage by 23.92%. The diff coverage is 95.95%.

@@             Coverage Diff             @@
##           master      #29       +/-   ##
===========================================
- Coverage   94.47%   70.55%   -23.93%     
===========================================
  Files          16       20        +4     
  Lines         163      180       +17     
===========================================
- Hits          154      127       -27     
- Misses          9       53       +44

Impacted Files	Coverage Δ
src/sauvola.jl	0% <ø> (-93.75%)	:arrow_down:
src/niblack.jl	0% <ø> (-93.75%)	:arrow_down:
src/ImageBinarization.jl	100% <ø> (ø)	:arrow_up:
src/polysegment.jl	0% <ø> (-91.67%)	:arrow_down:
src/minimum_error.jl	100% <100%> (ø)	:arrow_up:
src/intermodes.jl	100% <100%> (ø)	:arrow_up:
src/otsu.jl	100% <100%> (ø)	:arrow_up:
src/moments.jl	100% <100%> (ø)	:arrow_up:
src/BinarizationAPI/binarize.jl	100% <100%> (ø)
src/unimodal.jl	100% <100%> (ø)	:arrow_up:
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 2ff0e10...07704cb. Read the comment docs.

[codecov-io]

Codecov Report

Merging #29 into master will decrease coverage by 23.92%. The diff coverage is 95.95%.

@@             Coverage Diff             @@
##           master      #29       +/-   ##
===========================================
- Coverage   94.47%   70.55%   -23.93%     
===========================================
  Files          16       20        +4     
  Lines         163      180       +17     
===========================================
- Hits          154      127       -27     
- Misses          9       53       +44

Impacted Files	Coverage Δ
src/sauvola.jl	0% <ø> (-93.75%)	:arrow_down:
src/niblack.jl	0% <ø> (-93.75%)	:arrow_down:
src/ImageBinarization.jl	100% <ø> (ø)	:arrow_up:
src/polysegment.jl	0% <ø> (-91.67%)	:arrow_down:
src/minimum_error.jl	100% <100%> (ø)	:arrow_up:
src/intermodes.jl	100% <100%> (ø)	:arrow_up:
src/otsu.jl	100% <100%> (ø)	:arrow_up:
src/moments.jl	100% <100%> (ø)	:arrow_up:
src/BinarizationAPI/binarize.jl	100% <100%> (ø)
src/unimodal.jl	100% <100%> (ø)	:arrow_up:
... and 16 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 2ff0e10...88202fe. Read the comment docs.

[codecov-io]

Codecov Report

Merging #29 into master will increase coverage by 0.46%. The diff coverage is 96.69%.

@@            Coverage Diff             @@
##           master      #29      +/-   ##
==========================================
+ Coverage   94.47%   94.94%   +0.46%     
==========================================
  Files          16       20       +4     
  Lines         163      178      +15     
==========================================
+ Hits          154      169      +15     
  Misses          9        9

Impacted Files	Coverage Δ
src/util.jl	100% <ø> (ø)	:arrow_up:
src/ImageBinarization.jl	100% <ø> (ø)	:arrow_up:
src/minimum_error.jl	100% <100%> (ø)	:arrow_up:
src/intermodes.jl	100% <100%> (ø)	:arrow_up:
src/otsu.jl	100% <100%> (ø)	:arrow_up:
src/sauvola.jl	94.73% <100%> (+0.98%)	:arrow_up:
src/moments.jl	100% <100%> (ø)	:arrow_up:
src/BinarizationAPI/binarize.jl	100% <100%> (ø)
src/niblack.jl	100% <100%> (+6.25%)	:arrow_up:
src/unimodal.jl	100% <100%> (ø)	:arrow_up:
... and 14 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 2ff0e10...125c3e4. Read the comment docs.

[johnnychen94]

Regardless of all those dirty and lengthy details in this PR which I believe nobody would like to review, as a summary, this PR does:

• refactor the codebase using the functor API discussed in #26
• enhance the API by introducing a submodule BinarizationAPI
• add in-place function binarize!
• support Color3 inputs
• add more test codes
• slightly enhance the documentation

Breaking changes (Deprecated in 0.3):

• swap the argument order discussed in #23 ( https://github.com/zygmuntszpak/ImageBinarization.jl/pull/29/commits/d1f8309ef5fa35f46f52aa748e9de0307edd8fc6)
• unexport recommend_size in favor of #41, i.e., AdaptiveThreshold(img) instead of recommend_size(img) ( PR: #30 #45)
• made window_size of AdaptiveThreshold not an optional argument. ( PR: #45 )

made window_size of AdaptiveThreshold not an optional argument.

The idea here is to encourage users to use AdaptiveThreshold(img) to infer window_size rather than some fixed integer 32.

---

Any comments before I merge this PR? @zygmuntszpak