Noise Estimation and Type Identification in Natural Scene and Medical Images using Deep Learning Approaches

Algorithm 1

Proposed algorithm for noise estimation.

Input: Noised images

Output: Estimated noise level

Method:

Step 1: Read images

    For i = 1 to n do

       Read image I_i

       Repeat step 2 to 5 for each image

Step 2: Resize an image

       Resize image I_i to 256 × 256

Step 3: DWT transformation

       [LL, LH, HL, HH] = DWT transformation.

Step 4: Edge detection and reduce size to match with HH sub-band

       I_d = Detect edges of I_i using the Sobel operator.

       I_dd = Downsample the image by two rows and two columns.

Step 5: Remove edge component from HH sub-band

       I_we = HH ○ I_dd Hadamard operation

Step 6: Estimate noise level

       Initial noise level = Using Equation 3.

       Create a csv file with 2 column: estimated noise level and label.

Step 7: Train and test the neural fit model

    Divide csv file in the ratio 70 : 15 : 15.