Catalog · How the catalog works
Hough-Transformation
Hough transformation for circle and line detection
- ★ 1 on GitHub
Anyone may press Interesting, signed in or not. It helps others find this project.
README
# Hough-Transformation
Hough transformation for **circle** and **line** detection in Delphi / Pascal (FMX demo included)
Looking for the VCL version: https://sourceforge.net/projects/houghtransforma/ [2]
---
## Overview
This project contains a Delphi/Pascal implementation of the Hough Image Transformation algorithm.
It supports:
- **Line detection** via classic Hough space (θ, r) voting
- **Circle detection** via a circle Hough transform (radius provided by the user/UI) [1]
The implementation produces good results on simple test images and includes post-processing capabilities with dynamic thresholding [3].
---
## Algorithm Details
### Line Detection Algorithm
The line Hough transform works by:
1. **Voting in parameter space**: For each edge point in the image, the algorithm votes in a 2D accumulator array indexed by angle (θ) and distance (r) [3]
2. **Peak detection**: After voting, peaks in the accumulator represent likely lines in the original image
3. **Thresholding**: The implementation includes both fixed and dynamic thresholding methods to filter results [3]
The key mathematical relationship is:
- r = X \\\* cos(θ) + Y \\\* sin(θ)
The algorithm provides two thresholding approaches:
- **Fixed threshold**: Using HoughResultToParameter with a set range value [3]
- **Dynamic threshold**: Using HoughResultToParameterDynamic for adaptive filtering based on local statistics [3]
### Circle Detection Algorithm
Circle detection is exposed as a function that takes:
- An input bitmap
- A Hough accumulator/result structure
- A **test radius** (chosen by the user) [1]
The implementation is designed to run for a specific radius. To detect circles of unknown sizes, you would need to repeat the process across multiple radii [1].
---
## Project Structure
- Unit\\\_HoughTransformation.pas
Core algorithm implementation for line and circle detection, including helper functions for finding maxima and thresholding [3].
- GUI.FMX.HoughTransform\\\_Demo.pas
FireMonkey (FMX) demo GUI that loads an image, runs line/circle detection, and displays results [1].
- README.md
This file.
---
## API / Key Types
The algorithm unit defines:
- THoughResult = array of array of integer; - 2D accumulator array [3]
- THoughFinal - Structure containing filtered results with X, Y, and Z (vote count) coordinates [3]
Main entry points:
- procedure Hough\\\_LineDetection(AnalysisBitmap: TBitMap; var aHoughResult: THoughResult); [1]
- procedure Hough\\\_CircleDetection(AnalysisBitmap: TBitMap; var aHoughResult: THoughResult; Radius: Integer); [1]
Helper functions:
- function MAX\\\_2D(aHoughResult: THoughResult): integer; - Finds maximum value in accumulator [3]
- function MEAN\\\_2D\\\_Local(aHoughResult: THoughResult; x\\\_pos, y\\\_pos, x\\\_size, y\\\_size: integer): integer; - Calculates local mean [3]
- procedure HoughResultToParameterDynamic(aHoughResult: THoughResult; Range: Real; var aHoughFinal: THoughFinal); - Applies dynamic thresholding [3]
---
## How to Use
### 1) Line Detection (Basic)
var
\  AnalysisBitmap: TBitmap;
\  HoughResult: THoughResult;
\  ResultBitmap: TBitmap;
begin
\  // Prepare your source bitmap
\  AnalysisBitmap := TBitmap.Create;
\  // ... load image into AnalysisBitmap ...
\ 
\  // Run Hough line detection
\  Hough\\\_LineDetection(AnalysisBitmap, HoughResult);
\ 
\  // Convert result to viewable bitmap
\  ResultBitmap := TBitmap.Create;
\  HoughresultToBitMap(ResultBitmap, HoughResult);
\ 
\  // Display or save ResultBitmap
\  // ...
\ 
\  ResultBitmap.Free;
\  AnalysisBitmap.Free;
end;
2\\) Line Detection (With Thresholding)
```pascal
var
\  AnalysisBitmap: TBitmap;
\  HoughResult: THoughResult;
\  HoughFinal: THoughFinal;
\  ResultBitmap: TBitmap;
\  Threshold: Integer;
begin
\  // Run detection
\  Hough\\\_LineDetection(AnalysisBitmap, HoughResult);
\ 
\  // Apply dynamic thresholding
\  Threshold := 50; // Percentage (0-100)
\  HoughResultToParameterDynamic(HoughResult, (Threshold / 100), HoughFinal);
\ 
\  // Convert filtered results to bitmap
\  ResultBitmap := TBitmap.Create;
\  HoughResultLineParameterToBitMap(ResultBitmap, HoughFinal);
\ 
\  ResultBitmap.Free;
end;
3\\) Circle Detection
```pascal
var
\  AnalysisBitmap: TBitmap;
\  HoughResult: THoughResult;
\  ResultBitmap: TBitmap;
\  TestRadius: Integer;
begin
\  // Set the radius to search for
\  TestRadius := 25; // pixels
\ 
\  // Run Hough circle detection
\  Hough\\\_CircleDetection(AnalysisBitmap, HoughResult, TestRadius);
\ 
\  // Convert and display result
\  ResultBitmap := TBitmap.Create;
\  HoughresultToBitMap(ResultBitmap, HoughResult);
\ 
\  ResultBitmap.Free;
end;
FMX Demo Usage
The included FireMonkey demo provides a GUI for testing the algorithms
GUI.FMX.HoughTr...m\_Demo.pas
:
Loading an Image
Click the Load Image button
Select a test image from your file system
GUI.FMX.HoughTr...m\_Demo.pas
Line Detection
Load an image
Optionally adjust the Hough Threshold slider (controls filtering sensitivity)
GUI.FMX.HoughTr...m\_Demo.pas
Click Hough Line button to run line detection
GUI.FMX.HoughTr...m\_Demo.pas
View results:
Hough Result: Raw accumulator visualization
Hough Accumulator: Filtered line parameters overlaid on image
GUI.FMX.HoughTr...m\_Demo.pas
Circle Detection
Load an image
Set the radius value using the spinbox control
GUI.FMX.HoughTr...m\_Demo.pas
Click Hough Circle button to run circle detection
GUI.FMX.HoughTr...m\_Demo.pas
View the Hough Result image showing detected circles
GUI.FMX.HoughTr...m\_Demo.pas
Implementation Notes
Dynamic Thresholding
The implementation includes dynamic thresholding functionality that adapts to the accumulator data
Unit\_HoughTransformation.pas
:
Computes maximum values across the accumulator using MAX\_2D
Unit\_HoughTransformation.pas
Filters votes based on a percentage of the maximum (Range \* max)
Unit\_HoughTransformation.pas
Can incorporate local gradient information with configurable minimum gradient (Grad\_Min)
Unit\_HoughTransformation.pas
Result Interpretation
Bright pixels in accumulator = High vote counts = Likely line/circle parameters
The HoughFinal structure contains filtered (X, Y, Z) coordinates where:
X, Y = Parameter space coordinates (??, r for lines; center coordinates for circles)
Z = Vote count (confidence)