return to Smart Digital Platform home page
This is the app page of ODMO program for minimizing energy loss using GP
Otsu's method is used for separating foreground and background in an image by choosing an appropriate threshold of intensity. In the traditional method, the values of inter-class variance (between-class variance) are computed for all values of the thresholds, and then the optimal threshold is chosen according to the highest inter-class variance. If the number of gray levels and the number of images are numerous, then the computational complexity is extremely high.
Assume that the value of the initial threshold is T. The given gray image has L gray levels, N pixels, and a vector P = {P
_{0}
, P
_{1}
, ..., P
_{L-1}
} being the number of pixels in each gray levels. Some fundamental equations:
1. Background
2. Foreground
Weight: \({W_b} = \frac{\sum\limits_{i = 0}^{T - 1} {P_i} }{N}\)
Weight: \({W_f} = \frac{\sum\limits_{i = T}^{L - 1} {P_i} }{N}\)
Mean: \({\mu _b} = \frac{\sum\limits_{i = 0}^{T - 1} i{P_i} }{\sum\limits_{i = 0}^{T - 1} {P_i} }\)
Mean: \({\mu _f} = \frac{\sum\limits_{i = T}^{L - 1} i{P_i} }{\sum\limits_{i = T}^{L - 1} {P_i} }\)
The problem is: How to find the optimal threshold T more quickly?
Please upload your image (with 8-bit requirement) to the box and find out how quick our method is compared to the conventional method!!!
[CLICK] here to solve the optimization problem
Solve Problem
__________
__________