Computer vision as the name suggests the study of the visual system for artificially intelligent systems like robots. Computer vision involves using a combination of camera hardware and computer algorithms to allow robots to collect visual information from its environment.

Robotics is a field that is heavily inspired by nature. We can attribute the features of the robotic computer vision system to that of a human vision system. Humans collect visual information through their eyes which, is sent to the brain and processed for further work. Robotic vision also functions similarly. 

Basic pipeline:

Step 1: The robot scans the environment using a CCD (Charged Coupled Device) camera. 

Step 2: Light intensities cause the accumulation of varying charges at the photo-sites. An electron beam scanner is responsible for measuring these intensities in each direction. The analog plot of the light intensities obtained is digitized (A/D conversion).

Step 3: The image is stored in the memory in the form of an array. This step is also known as the Frame Grabbing step.

Step 4: The obtained image information is operated upon to achieve a better vision. Techniques to achieve noise reduction and lost information retrieval are performed.

In this article, We shall cover the most basic and first block in the study of machine computer vision. Let us begin by introducing ourselves to the idea behind a pinhole camera and photometry.


The human eye is not equally sensitive to all wavelengths of visible light. Radiometry is the science of measuring light while photometry concentrates on the visible areas of light. Photometry attempts to account for this by weighing the measured power at each wavelength with a factor that represents how sensitive the eye is at that wavelength. 

So how is the brightness at a pixel in the scene image determined?

 To answer this question let us first understand the concept of Radiance, Irradiance, and Solid Angle.


Radiance is the power per unit foreshortened area emitted into a unit solid angle (W · m−2 sr−1 watts per square meter per steradian) by a surface. Luminance is a spectrally weighted radiance (Lumen · m−2· sr−1).


Irradiance is the power per unit area (W · m−2 watts per square meter) of the radiant energy falling on a surface. Illuminance is spectrally weighted irradiance (Lumen · m−2).

Solid Angle:

The solid angle of a cone of directions is defined as the area cut out by the cone on the unit sphere. It is formulated as $\Omega=\frac{A \cos \theta}{R^{2}}$. Using simple logic we know that the more the power radiated by the scene the brighter the image will be. Therefore,

Image irradiance ${\displaystyle \propto }$ Scene radiance

We will revisit the topic of brightness. For now, I leave you to think about how the solid angle, the power reflected by the scene, the power received by the scene, and the above proposition can be related to output the image irradiance. Below is the template you can brainstorm with.

pinhole camera basic template

Pinhole Camera:

Let us now try to capture an image from the given scene. 

computer vision peacock

The image shows a peacock standing against the images sensors (which are in reality inside the camera). If you notice all the image sensors are equally exposed to each part of the peacock. This will cause each image sensor to pick up rays from each reflecting point the peacock’s body and cause the image to appear blurry as shown in the below photo.

computer vision peacock

To avoid this let us introduce a slit layer between the sensors and object of interest. This will each sensor to pick energy from only a certain localised region. For a better understanding look at the image below.

pinhole lens

The image obtained is dark as the energy collected from a particular point on the image sensor is substantially less in amount. Therefore we get the following as a result of this.

inverted pinhole

We are getting a better image but need to collect more energy so that the image appears brighter. We do that using a lens as is shown below. The lens will be able to focus more energy on the image sensors.

computer vision peacock

This is the basic concept behind a pinhole camera. You should try to make one at home too for a better understanding. You can make one using the instruction given in the video below.



I hope you liked this basic lesson on robotic computer vision. This will help you to get started in the field. The material has been referenced from the book Robot vision. MIT Press, 1986. The pinhole example images have been referenced from the slides of Dr. Rajendra Nagar taking the course on digital image processing at IIT Jodhpur. You may also want to check out my article on machine translation here. Thank you.

Looking for projects in computer vision ??Here is a list of 500+ project ideas compiled for you Top 25 computer vision ideas compiled 2020.

No Thoughts on Computer Vision: Starting with a pinhole camera

Leave A Comment