OpenCV calcHist function.

Question

I am using the calcHist function to calculate color histograms of images. I am working on a simple image matching application and I am trying to use color histograms as image descriptor to find similar images.

This is the definition:

void calcHist(const Mat* images, int nimages, const int* channels, InputArray mask, OutputArray hist, int dims, const int* histSize, const float** ranges, bool uniform=true, bool accumulate=false )

My question is regarding the dims parameter. This is from the OpenCV documentation:

dims – Histogram dimensionality that must be positive and not greater than CV_MAX_DIMS (equal to 32 in the current OpenCV version).

This might be a dummy question, but what is the difference between the results if I set dims as 1 or 3? If I set dims=1 am I only getting histogram for a single channel in the image or is this a histogram that represents all channels in 1D vector?

Micka Micka · Accepted Answer · 2015-03-04T16:02:09

Example:a 4x4 HS(V) Image:

Hue channel:

1   170 33  14
15  133 33  7
122 90  100 142
1   170 33  14

Saturation channel:

25  33  33  45
25  55  255  255
25  33  255  255
10  55  255  255

now assume 3 histogram BINS per channel:

HA = [0..60]
HB = [60..120]
HC = [120..180] 
 for H channel

SA = [0..80]
SB = [80..160]
SC = [160..255] 
 for S channel

now the difference:

dim = 1 would make a 1D histogram for each channel:

#HA = 9
#HB = 2
#HC = 5

and

#SA = 10
#SB = 0
#SC = 6

if you use dim = 2 you get the inter-channel histogram:

-  HA HB HC
SA  5  1  4  
SB  0  0  0
SC  4  1  1

which is a more detailed information.

Hope this is the answer to your question.

OpenCV calcHist function.

2 Answers