Adaptive Bitrate (ABR) streaming optimizes for individual
Quality of Experience (QoE) under variable network
conditions. Nevertheless, this isolated approach often
leads to systematic unfairness and resource contention when
multiple clients share a bottleneck link. To address this,
we propose Cooperative MPC (Model Predictive Control), a
modified model predictive control formulation that applies
an epidemic gossip protocol to share throughput
observations among clients. By limiting communication to a
small, constant number of random peers per gossip round,
the protocol aims to achieve swarm-wide state convergence
with overhead that scales independently of the total client
population. In this paper we evaluate this approach through
a proof-of-concept simulation across varying swarm sizes up
to 128 concurrent clients. Initial results indicate that
cooperative throughput-sharing can effectively mitigate the
aggressive spike exploitation inherent in standalone MPC
algorithms, thereby stabilizing bitrate selection and
reducing rebuffering events. Furthermore, the protocol
demonstrates resilience to realistic message delays and
extended gossip periods, suggesting potential viability for
deployment at CDN (Content Delivery Network) edges or dense
local networks.

Existing synthetic aperture radar (SAR) imagers produce a
huge amount of data that can be useful for various
applications. Many SARs are installed on-board of airborne
or spaceborne carriers and, then, acquired data or images
have to be transferred from on-board sensors and processors
to on-land centers of further data processing. In such
cases, data compression is extremely useful where the use
of lossy compression is preferable. Acquired SAR images are
contaminated by speckle and this leads to specific effects
of speckle suppression and possible existence of optimal
operation point (OOP) where compressed image is closer to
true image than uncompressed image according to certain
metrics. Compression performance depends on several factors
including a used coder. Recently Versatile Video Coding
(VVC) technique has been put forward that can be also
applied to still images. Thus, our goals are to consider
whether or not VVC can be used for SAR image lossy
compression, does OOP exist and, if yes, what are coder
parameters in OOP. We show that OOP might exist according
to compression ratio (CR), different full-reference (FR)
and no-reference (NR) metrics, and this happens for quite
large CRs. For real-world applications, the OOP position,
influenced by speckle intensity, can be estimated using NR
metrics or derived from the CR curve, while FR metrics
serve as a benchmark under laboratory conditions. The
corresponding practical recommendations are given.

This paper presents an enhanced registration algorithm
integrated with a hybrid metaheuristic optimization method,
DE-PSO (Differential Evolution-Particle Swarm
Optimization). By combining the global exploration
capability of DE with the fast convergence characteristics
of PSO, the proposed approach effectively overcomes the
local minima limitations commonly encountered in
conventional ICP implementations. The proposed method is
implemented entirely using free and open-source software
and developed in the C++ programming language on a
self-generated dataset. Experimental results demonstrate
that incorporating the DE-PSO optimization strategy within
the registration process significantly improves alignment
accuracy, robustness, and overall performance compared to
traditional registration approaches.

Automation in the video surveillance field has gained
significant traction among researchers. The primary
objective is to replace traditional human-dependent
monitoring with new solutions driven by robust computer
vision algorithms. This paper analyzes methods relevant for
Human Activity Recognition (HAR) and movement analysis
across three distinct scenarios: crowd analysis, sparse
object detection, and individual tracking. By integrating
CSRNet (Congested Scene Recognition Network) for crowd
counting, YOLOv12 (You Only Look Once) for sparse object
detection, and IMU-based (Inertial Measurement Unit)
individual activity monitoring, we derive multi-stage time
series. Given that each stage introduces varying
computational complexity, a preliminary analysis of
inference latency is presented. The results demonstrate
stage-specific performance and offer preliminary insights
towards minimizing latency during the monitoring process.

Road surface profile quality plays an important role in
defining pavement material functionality. Laser profiling
systems are valuable tools for traffic safety, particularly
in relation to friction effects. The modelling process
begins with contactless remote surface scanning of
materials. It is based on the interaction between coherent
light and the surface, followed by various
signal-processing techniques. Mobile scanning systems are
based on multisensory platforms moving along the road
surface. It is important to evaluate movement-related
effects such as vibrations of the moving system, changes in
platform velocity, macro-level surface irregularities, and
similar structural material issues.
Precise laser signal scanning techniques contribute to the
development of risk prediction models that support traffic
safety in situations where friction-related problems arise.
Moreover, these mobile profiling methods are
cost-effective. They improve safety, maintenance planning,
and lifecycle cost management while reducing both survey
time and traffic disruption without requiring road lane
closures.

U ovom radu izvršena je komparativna analiza JPEG, BPG i
VVC tehnika kompresije na slikama sistema daljinskog
osmatranja. Korišćena je baza slika DOTA v1.0, koja sadrži
snimke sa satelita GF-2 i JL-1, kao i sa platforme Google
Earth. Nakon podele originalnih slika na blokove dimenzija
1024x1024 piksela, formiran je skup od 15749 slika, od
kojih je 90 nasumično odabrano za testiranje. Kvalitet
komprimovanih slika procenjen je korišćenjem tri objektivne
mere procene kvaliteta, dok je efikasnost kompresije
izražena preko reprezentacije u bitima po pikselu i stepena
kompresije. Poređenje je izvršeno ujednačavanjem
objektivnog kvaliteta i ujednačavanjem broja bita po
pikselu. Rezultati pokazuju da BPG i VVC ostvaruju dva do
pet puta veću uštedu memorijskog prostora u odnosu na JPEG
za isti nivo kvaliteta. Pri istom broju bita po pikselu,
BPG i VVC daju bolji objektivni kvalitet, sa razlikama u
vršnom odnosu signal šum do 7 dB. Takođe je uočeno da tip
degradacije različito utiče na pojedine mere kvaliteta.

Rad predstavlja analizu uticaja kompresije slike sa
gubicima na sakrivanje informacija u prostornom domenu
metodom zamene bita najmanjeg značaja. Performanse
steganografske tehnike određivane su korišćenjem vršnog
odnosa signal-šum. Akcenat je dat na metode kompresije
slike sa gubicima, ali je izvršena analiza i za slučaj kada
su primenjene tehnike kompresije bez gubitaka. Utvrđeno je
da je u oba slučaja osobina neprimetnosti na
zadovoljavajućem nivou. Primenom kompresije sa gubicima na
steganografske slike dolazi do značajnog oštećenja
sakrivene informacije. Pokazano je da je JPEG2000 tip
kompresije najotporniji na oštećenja sakrivene poruke. BPG
tip kompresije se pokazao kao najefikasniji sa stanovišta
uštede memorije, naročito kada je u pitanju kompresija bez
gubitaka.