* Correction for computeBinaryMap crash:
- the function staticSaliencyFineGrained::computeSaliency() outputs a
normalized floating point image
- Add an assert in StaticSaliency::computeBinaryMap in order to check
image depth (must be CV_32F)

* - Change assert in cv::Saliency::StaticSaliency::computeBinaryMap()
- Change normalization factor in cv::Saliency::StaticSaliencyFineGrained::computeSaliencyImpl()

In some case, Discrete Fourier Transform in computeSaliencyImpl() returns
magnitude matrix which contains zero values.
Then, log() returns -inf values and normalization with blur() returns -nan.
When computeBinaryMap() is called double free or corruption exception occurs
because kmeans() fails to compute distance.
Signed-off-by: d.bouron <d.bouron@criteo.com>

Old one causes build errors.

noise control implementation ended

noise control errors fixed (part I)

noise control errors fixed (ended)

Perf Test for staticSaliencySpectralResidual and objectnessBING added

improvements and bugs fixed

saliency perf test added

matrices optimization

Where possible, CV_8U matrices have been used in place of CV_32F

new sorting algorithm implemented (first part)

new sorting algorithm implemented (second part)

new sorting algorithm implemented (finished)

- made some of dependencies explicit
- removed dependencies to highgui and some other modules where possible
- modified some samples to build without modules

Fixing compilation

Patched the saliency module to use ObjectivenessBING bounding box
proposals from python.

Usage:

```
bing = cv2.saliency.ObjectnessBING()
img = cv2.imread('/home/ignotus/Pictures/10376146_973700932714682_2056967515849182952_n.jpg')
bing.computeSaliency(img)
```

The rectangle at top is not having maximum objectness value.

Fixed warnings for XCode 7.1.1 and cmake 3.3.2, rewrote one method in sliency to avoid clang 7.0.0 crash

Adding float casts allows them to compile on MSVC and should not
affect other compilers.

NAN is a compiler specific constant that does not exist on
visual studio. I also noticed that some NAN instances had been
cast to float, whereas quiet_NaN is both 1) defined by the
standard and 2) templated for the correct type. This is a much
more portable method of getting a NaN value.

replaced "const InputArray" with "InputArray" to avoid warnings about "const const _InputArray&"