# UK EE 630 - EE630 PROJECT A STRIPE DETECTION AND SEGMENTATION (8 pages)

Previewing pages 1, 2, 3 of 8 page document
View Full Document

## EE630 PROJECT A STRIPE DETECTION AND SEGMENTATION

Previewing pages 1, 2, 3 of actual document.

View Full Document
View Full Document

## EE630 PROJECT A STRIPE DETECTION AND SEGMENTATION

25 views

Problems/Exams

Pages:
8
School:
University of Kentucky
Course:
Ee 630 - Digital Signl Processing
##### Digital Signl Processing Documents
• 9 pages

• 7 pages

• 2 pages

• 2 pages

• 3 pages

• 2 pages

• 7 pages

Unformatted text preview:

EE630 PROJECT A STRIPE DETECTION AND SEGMENTATION by LGH Updated 10 15 12 You will be graded on performing steps and answering questions indicated by 1 INPUT DATA Given a bmp image HOLDPATFRED bmp Figure 1 Mannequin Fred with triangular stripe patter projected at an angle from the camera s optical axis Step 1 The image can be brought into MATLAB with the following code A bmp double imread HOLDPATFRED bmp load HOLDPATFRED jpg image Ar A bmp 1 Ag A bmp 2 Ab A bmp 3 The image is B W so the Ar Ag Ab To plot Ar use imagesc Ar followed by colormap gray Step 1 1 Plot image of input data file 2 SELECT SIGNAL COLUMN 1 Step 2 1 Select and plot the nx column to represent a signal Indicate which nx you used The size of the image is My Nx size Ar A 1 D vector would be My by 1 sc Ar 1 My nx Figure 2 The middle column intensity of Fig 1 3 FIND THE SPECTRA OF THE COLUMN Step 3 1 Take the DFT of Figure 2 and plot as in Figure 3 such that Figure 3 FFT of Figure 2 with fftshift to center the dc term 2 Questions 3 1 Knowing the vertical height of the image units of pixels in Fig 1 estimate the number of stripe cycles occurring along the vertical direction of the image 3 2 Does this value correspond to the peak locations in Fig 3 3 3 Which peaks 4 STRIPE SEGMENTING We would like to segment the striping from the image of Fred To do this we can use an ideal bandpass filter An ideal bandpass filter is two rectangle functions symmetric about dc in the frequency domain To create a bandpass filter we can use a discrete time cosine function and the downloadable irect The discrete cosine can be synthesized as t k 1 kc c cos 2 pi kc t My c c Combining this with the rectangle function H irect 1 41 1 My H H h ifft H time domain DT Sinc such that the bandpass filter is Hb fft c h Step 4 1 generate the bandpass filter and manually optimize the center frequency Plot results as in Figs 4 and 5 The plot of the filter superimposed on the spectra is Figure 4 Bandpass filter superimposed on the signal

View Full Document

Unlocking...