How to cite self-created images or pictures in thesis
There is no need to cite self-created, previously-unpublished images any more than you need to cite self-written, previously unpublished words. By absence of a citation you are claiming the work as your own.
Since this is for a thesis, you should probably check with your advisor / supervisor to be absolutely certain they share that view.
Have a look at the relevant guidelines/rules at your university/department.
Where I am, one typically signs that everything that is not your work or that is not novel is cited. Novelty in the thesis context is defined as "has not been used in any other exam/thesis". Papers published as part of the thesis work do not hamper thesis-relevant novelty at "my" university. But since it is up to the unviersity (or even faculty) to decide their exam rules, you may be subject to different rules.
In addition, copyright plays a role. You may be author without having the copyright. The holder/owner of the copyright can allow you to re-use images but stipulate certain ways of citing.
Anything (images, diagrams, tables, text) you created for the current thesis (so they are novel according to the thesis rules) and where you hold the copyright (so no rules binding to use particular citation forms) are not cited. If there is no citation, you claim it's yours and novel for this thesis - and thus that it should be included in the content based on which your work is judged by the thesis committee.
In my case, I needed to cite my own work only if/when:
I wanted to reuse an image/diagram that I used already for an earlier thesis or exam, e.g. had I wanted a diagram from my Master thesis to re-appear in my PhD thesis. In that case, I'd have cited my Master thesis like unrelated work. Unrelated since this is not part of the work the committee should judge.
I transfered copyright for several papers I authored to the respective publishers. They typically did allow re-use in theses but prescribed citation sentences like e.g. "This image is reused from [CB3] with kind permission of
$publisher." [CB3] would be how I cited e.g. the third paper I wrote as part of PhD current thesis.
Had I retained copyright (non-exclusive license to the publisher), I wouldn't have need to cite this way. Nevertheless, I'd still think it advisable to tell the reader that this diagram can also be found in that paper - it saves people from trying to find out whether the paper contains anything in addition to what the thesis contains. Similarly, I could have used [3], i.e. there was no requirement to make the citiations of my papers for the thesis immediately distinguishable from other citations, but I thought it a good idea to provide such a distincition - e.g. because it makes it clear to the committee that this is part of what they judge as my thesis.
For your scenario of diagrams illustrating some application architecture described in some paper:
if the diagram for the existing one is directly taken from the paper, you apply for a license to re-use that image with the publisher and then do whatever they stipulate in this license.
if you draw your own diagram (in your architecture diagram "language"/style) from the description in the paper, you cite the idea/description, e.g. "diagram of architecture described in [27]" but the diagram itself is your own work, and novel and original part of your thesis.
With "your own diagram" I mean e.g. situations like: They have some diagram in the paper. But their way of illustrating isn't really suitable for your thesis since you need to highlight/contrast aspects that are similar to|differ from the architecture you developed. You therefore make a new graph that describes their architecture in another language/style that allows to depict their architecture as well as your architecture.
Your diagram of your architecture is original and novel work that is part of your thesis. No citation here, unless you signed away the copyright for this diagram - in which case you need a license by the new copyright owner (= journal publisher). In which case you do whatever the license says you should do.