EP06_08 — ⚫ QRCode Segmentation and Real Decoding with OpenCV

6.14.8.1 📋 Implementation Guidelines

1. Reading dimensions and parameters

   Read, in this order, from standard input:

   • one line containing the number of rows \(L\);
   • one line containing the number of columns \(C\);
   • one line containing the four algorithm parameters separated by spaces:
     • binarization threshold \(T\) (integer);
     • minimum area \(A_{\text{min}}\) (integer);
     • aspect tolerance \(\text{tol}\) (real);
     • margin \(M\) (integer, in pixels).

2. Image loading

   Use the didactic function f = mm.readImg(L, C) to read the \(L \times C\) values of the grayscale image, obtaining a NumPy array of type uint8.

3. Binarization

   Apply inverted binary thresholding using the threshold \(T\). Every pixel of the original image with intensity strictly greater than \(T\) must be converted to 255, while the remaining ones must assume the value 0.

4. Contour detection

   Extract the external connected components using cv2.findContours(...) with the parameters:

   • cv2.RETR_EXTERNAL;
   • cv2.CHAIN_APPROX_SIMPLE.

5. Geometric filtering

   For each contour found:

   • compute the bounding rectangle (x, y, w, h) using cv2.boundingRect;

   • keep only candidates that simultaneously satisfy:

     Minimum area

     \[ w \times h > A_{\text{min}} \]

     Aspect ratio

     \[ \left|\frac{w}{h}-1\right| \le \text{tol} \]

6. Candidate ordering

   Sort the candidates by the bounding rectangle area

   \[ w \times h \]

   in descending order.

   In case of a tie, preserve the order originally returned by cv2.findContours.

7. Verification by decoding

   For each candidate, following the established order:

   • expand the rectangle by \(M\) pixels in all four directions;
   • clamp the indices to remain within the image;
   • extract the crop directly from the original image f;
   • apply cv2.QRCodeDetector().detectAndDecode(...) to that crop.

8. Stopping criterion

   Immediately stop processing when the first candidate yields a non-empty decoded string.

9. Case not found

   If no candidate is successfully decoded, print exactly: QRCODE_NAO_ENCONTRADO

10. Output (found case)

    Print two lines.

    First line: linha coluna altura largura using the original bounding rectangle, before the expansion by the margin \(M\).

    Second line: texto_decodificado

6.14.8.2 📌 Computational Restrictions

• Use OpenCV functions to perform binarization, contour detection, bounding rectangle computation, and QRCode decoding.
• Geometric filtering must necessarily occur before the decoding step.
• Use exclusively the fixed threshold \(T\) provided in the input. It is not allowed to use automatic thresholding methods, such as Otsu or adaptive thresholding.
• Ensure that the crops sent to the decoder remain within the image boundaries.

6.14.8.3 🧠 Theoretical Foundation

This exercise integrates concepts studied throughout the chapter into a single Computer Vision pipeline. Segmentation reduces the set of candidate regions through geometric characteristics, while the decoding step validates the content of the region using a specialized recognition algorithm.

6.14.8.4 📦 Input and Output Specification (VPL)

Input Structure

L
C
T A_min tol M
[image matrix]

Output Structure (Success)

linha coluna altura largura
texto_decodificado

Output Structure (Failure)

QRCODE_NAO_ENCONTRADO

6.14.8.5 📌 Reference Files (.pgm)

For validation purposes, local debugging, and analysis of real pixel matrices, the image files generated in ASCII P2 format are available in the project directory. You may use them to test the adherence of your code by decoding them with your cell phone (save the *.pgm files locally to view them):

• 📥 Case 1: Normal Pattern – Contains a single perfectly centered code with simple geometric distractors on the periphery.
• 📥 Case 2: Complex Scenario – Presents a higher density of textured noise and multiple candidate distractors that test the limits of aspect filtering.
• 📥 Case 3: Expanded Message – Contains a QRCode structured from a longer character string, generating a higher density of internal modules.
• 📥 Case 4: Compact Geometry – Evaluates the pipeline behavior under optimized contrast conditions and borderline positioning.
• 📥 Case 5: Exclusion Scenario – An image composed purely of high-area distractor elements, designed to validate the controlled failure behavior of the program.

%%writefile EP06_08.py
# Python Code

Overwriting EP06_08.py

TestSuite("EP06_08.py").run()

✔️ EP06_08.cases already exists in casos/
📋 5 case(s) loaded from casos/EP06_08.cases

🔍 Testing Python: EP06_08.py
⚠️ EP06_08.py: Empty file (fewer than 3 lines). Tests skipped.