phovision 0.1.5

A pure Python computer vision library

Phovision

A pure Python computer vision library implementing various image processing algorithms from scratch. This library provides implementations of common image processing operations with no dependencies except NumPy.

Features

Image Processing:

• Gaussian Blur
• Median Filter (Noise Removal)
• Mean Filter (Averaging)
• Bilateral Filter (Edge-preserving smoothing)

Image I/O (Pure Python Implementation):

• Read images from files, URLs, base64 strings, or bytes
• Save images in various formats
• Convert images to base64 strings
• Supported formats: PNG (read), BMP (read/write), PPM/PGM (read/write)

Installation

pip install phovision

Usage

Basic Image Processing

from phovision import (
    read_image, save_image,
    gaussian_blur, median_filter, 
    mean_filter, bilateral_filter
)

# Read an image
image = read_image('input.bmp')  # or .ppm, .pgm, .png

# Apply filters
blurred = gaussian_blur(image, kernel_size=5, sigma=1.0)
denoised = median_filter(image, kernel_size=3)
averaged = mean_filter(image, kernel_size=3)
smoothed = bilateral_filter(image, kernel_size=5, sigma_spatial=1.0, sigma_intensity=50.0)

# Save results
save_image(blurred, 'blurred.bmp')  # or .ppm, .pgm
save_image(denoised, 'denoised.bmp')

Flexible Image I/O

from phovision import read_image, save_image, to_base64

# Read from local file
img1 = read_image('local_image.bmp')

# Read from URL
img2 = read_image('https://example.com/image.png')

# Read from base64 string
img3 = read_image('data:image/bmp;base64,...)

# Convert to base64 (useful for web applications)
base64_str = to_base64(img1, format='PPM')  # or 'BMP'

# Save in different formats
save_image(img1, 'output.bmp')   # Windows Bitmap
save_image(img1, 'output.ppm')   # Portable Pixmap (color)
save_image(img1, 'output.pgm')   # Portable Graymap (grayscale)

Working with NumPy Arrays

import numpy as np
from phovision import read_image, gaussian_blur

# Create a synthetic image
width, height = 200, 200
x, y = np.meshgrid(np.linspace(0, 1, width), np.linspace(0, 1, height))
synthetic = np.sin(10 * x) * np.cos(10 * y) * 127 + 128
synthetic = synthetic.astype(np.uint8)

# Process the synthetic image
processed = gaussian_blur(synthetic, kernel_size=5, sigma=1.0)

# The library works directly with numpy arrays
if isinstance(processed, np.ndarray):
    print("Output is a numpy array!")

Requirements

• Python >= 3.7
• NumPy >= 1.21.0

License

MIT License