To address the questions based on the provided image:

(i) By observation, identify the type of transformation that exists between the images.

Observing the lines connecting key points between the two images, we notice that:

1. There is a consistent alignment and direction in the matched points, but the points are not identical.
2. The points appear to have both a translation and a slight change in orientation (rotation), with minimal to no scaling effect.

Based on this observation, the transformation is likely an affine transformation. An affine transformation includes translation, rotation, and possibly some scaling or shearing, but it preserves parallel lines.

(ii) List steps that must be taken to compute the above transformation between the two images.

To compute the transformation, you would typically follow these steps:

1. Detect Key Points: Use a feature detector algorithm (like SIFT, SURF, or ORB) to identify key points in both images.
2. Match Key Points: Find corresponding key points between the two images. This is often done using a feature matching algorithm (e.g., brute-force matching, FLANN).
3. Select a Transformation Model: Since an affine transformation is suspected, use an affine transformation model. Alternatively, RANSAC (Random Sample Consensus) can be applied to eliminate outliers and find the best transformation.
4. Estimate Transformation Parameters: Calculate the parameters of the affine transformation (translation, rotation, and any scaling/shearing if present) based on the matched points.
5. Apply the Transformation: Use the computed transformation matrix to align or transform the second image to the coordinate system of the first image.

Would you like further details on any of these steps, or do you have additional questions?

Related Questions:

1. What is the difference between affine and homography transformations?
2. How does RANSAC help in refining transformations?
3. What are the limitations of affine transformations in image matching?
4. Can affine transformations handle perspective distortions?
5. How are scaling and rotation parameters computed in an affine transformation?

Tip: For image matching tasks, robust feature detection methods (like SIFT or ORB) can greatly improve the accuracy of transformation estimation.